WEBVTT

00:00:00.080 --> 00:00:04.797
The following is a conversation with
Mark Andre, co-creator of Mosaic, the

00:00:04.880 --> 00:00:08.957
first widely used web browser,
co-founder of Netscape, co-founder of

00:00:09.040 --> 00:00:10.797
the legendary Silicon Valley venture

00:00:10.880 --> 00:00:14.237
capital firm Andre Horowitz, and is one

00:00:14.320 --> 00:00:16.317
of the most outspoken voices on the

00:00:16.400 --> 00:00:19.117
future of technology, including his most

00:00:19.200 --> 00:00:22.477
recent article, Why AI will save the

00:00:22.560 --> 00:00:26.397
world. And now,
a quick few second mention of

00:00:26.480 --> 00:00:29.917
each sponsor. Check them out in the
description. and it's the best way to

00:00:30.000 --> 00:00:33.997
support this podcast. We've got Inside
Tracker for tracking your health,

00:00:34.080 --> 00:00:36.637
ExpressVPN for keeping your privacy and

00:00:36.720 --> 00:00:39.917
security on the internet, and AG1 for my

00:00:40.000 --> 00:00:44.637
daily multivitamin drink. Choose wisely,
my friends. Also, if you want to work

00:00:44.720 --> 00:00:48.557
with our amazing team, we're always
hiring. Go to lexfreedman.com/hiring.

00:00:48.640 --> 00:00:52.877
And now on to the full ad reads.
As always, no ads in the middle. I try to

00:00:52.960 --> 00:00:56.557
make this interesting, but if you
skip them, please still check out our

00:00:56.640 --> 00:01:00.797
sponsors. I enjoy their stuff.
Maybe you will, too.

00:01:00.880 --> 00:01:03.277
This show is brought to you by Inside

00:01:03.360 --> 00:01:06.077
Tracker, a service I use to track

00:01:06.160 --> 00:01:07.837
whatever the heck is going on inside my

00:01:07.920 --> 00:01:11.597
body using data, blood test data that

00:01:11.680 --> 00:01:15.677
includes all kinds of information. And
that raw signals, processes, and machine

00:01:15.760 --> 00:01:19.357
learning to tell me what I need to do
with my life, how I need to change,

00:01:19.440 --> 00:01:21.517
improve my diet, how I need to change,

00:01:21.600 --> 00:01:23.757
improve my lifestyle, all that kind of

00:01:23.840 --> 00:01:26.637
stuff. I'm a big fan of using as much

00:01:26.720 --> 00:01:30.557
raw data that comes from my own body

00:01:30.640 --> 00:01:33.677
processed through generalized uh machine

00:01:33.760 --> 00:01:37.037
learning models to give a prediction to

00:01:37.120 --> 00:01:41.037
give a suggestion. This is obviously the
future and the more data the better and

00:01:41.120 --> 00:01:46.557
so companies like inside track doing an
amazing job of taking the leap into that

00:01:46.640 --> 00:01:50.237
world of personalized data and
personalized datadriven suggestion. I'm

00:01:50.320 --> 00:01:52.957
a huge supporter of it. Turns out that

00:01:53.040 --> 00:01:56.157
luckily I'm pretty healthy, surprisingly

00:01:56.240 --> 00:02:00.397
so. But then I look at the life and the

00:02:00.480 --> 00:02:03.837
limb and the health of uh Sir Winston

00:02:03.920 --> 00:02:07.117
Churchill who probably had the

00:02:07.200 --> 00:02:11.437
unhealthiest sort of diet
and lifestyle of any human

00:02:11.520 --> 00:02:13.837
ever and lived for quite a long time and

00:02:13.920 --> 00:02:16.797
as far as I can tell was quite nimble

00:02:16.880 --> 00:02:20.237
and agile into his old age. Anyway, get

00:02:20.320 --> 00:02:25.240
special savings for a limited time
when you go to insettracker.com/lex.

00:02:25.680 --> 00:02:27.197
This show is also brought to you by

00:02:27.280 --> 00:02:29.517
ExpressVPN. I use them to protect my

00:02:29.600 --> 00:02:32.717
privacy on the internet. It's the first

00:02:32.800 --> 00:02:36.077
layer of protection in this dangerous

00:02:36.160 --> 00:02:38.477
cyber world of ours that soon will be

00:02:38.560 --> 00:02:41.437
populated by humanlike or superhuman

00:02:41.520 --> 00:02:44.877
intelligent AI systems that will trick

00:02:44.960 --> 00:02:49.037
you and try to get you to do all kinds
of stuff. It's going to be a wild wild

00:02:49.120 --> 00:02:53.117
world in the 21st century. Cyber security,
the attackers, the defenders.

00:02:53.200 --> 00:02:55.037
It's going to be a tricky world. Anyway,

00:02:55.120 --> 00:02:56.797
a VPN is a basic shield you should

00:02:56.880 --> 00:03:00.637
always have with you in this battle for

00:03:00.720 --> 00:03:05.037
privacy, for security, all that kind of
stuff. What I like about it also is that

00:03:05.120 --> 00:03:09.197
uh it's just a well implemented piece
of software that's constantly updated. It

00:03:09.280 --> 00:03:13.517
works well across a large number of
operating systems. It does one thing and

00:03:13.600 --> 00:03:17.677
it does it really well. I've used it
for many, many years. Before I had a

00:03:17.760 --> 00:03:19.917
podcast, before they were a sponsor, I

00:03:20.000 --> 00:03:23.197
have always loved ExpressVPN with a big

00:03:23.280 --> 00:03:27.117
sexy button that just has a power symbol.
You press it and it turns on.

00:03:27.200 --> 00:03:30.877
It's beautifully simple.
Go to expressvpn.com/lebod

00:03:30.960 --> 00:03:33.917
for an extra 3 months free.

00:03:34.000 --> 00:03:35.757
This show is also brought to you by

00:03:35.840 --> 00:03:39.917
Athletic Greens and it's AG1 drink. It's

00:03:40.000 --> 00:03:42.317
an all-in-one daily drink to support

00:03:42.400 --> 00:03:43.917
better health and peak performance. I

00:03:44.000 --> 00:03:47.357
drink it at least twice a day now in the

00:03:47.440 --> 00:03:49.997
crazy Austin heat. It's over 100° for

00:03:50.080 --> 00:03:53.037
many days in a row. There's few things

00:03:53.120 --> 00:03:55.117
that feel as good as coming home from a

00:03:55.200 --> 00:03:59.117
long run and uh making it an age one

00:03:59.200 --> 00:04:03.117
drink, putting it in the fridge so it's
nice and cold. I jump in the shower,

00:04:03.200 --> 00:04:05.757
come back, drink it, I'm ready to take

00:04:05.840 --> 00:04:08.637
on the rest of the day.

00:04:08.720 --> 00:04:13.037
I'm kicking ass, empowered by the
knowledge that I got all my vitamins and

00:04:13.120 --> 00:04:17.037
minerals covered.
It's the foundation for all the wild

00:04:17.120 --> 00:04:19.517
things I'm doing mentally and physically

00:04:19.600 --> 00:04:24.877
with the rest of the day. Anyway,
they'll give you a onemon supply of fish

00:04:24.960 --> 00:04:29.240
oil when you sign up at drinkagg1.com/lex.

00:04:29.600 --> 00:04:33.480
That's drinkaggone.comx.

00:04:36.000 --> 00:04:39.277
This is the Lex Freedman podcast.
To support it, please check out our

00:04:39.360 --> 00:04:41.677
sponsors in the description. And now,

00:04:41.760 --> 00:04:46.920
dear friends, here's Mark Andre.

00:04:54.100 --> 00:04:56.120
[music]

00:05:01.120 --> 00:05:04.317
I think you're the right person to talk
about the future of the internet and

00:05:04.400 --> 00:05:07.197
technology in general. Um, do you think

00:05:07.280 --> 00:05:09.757
we'll still have Google search in five

00:05:09.840 --> 00:05:12.797
in 10 years or search in general?

00:05:12.880 --> 00:05:16.637
>> Yes. You know, it'll be a question if
the use cases have really narrowed down.

00:05:16.720 --> 00:05:19.517
Well, now with the AI, >> yeah,

00:05:19.600 --> 00:05:22.317
>> and AI assistance
being able to interact

00:05:22.400 --> 00:05:25.597
and expose the entirety of human wisdom

00:05:25.680 --> 00:05:27.437
and knowledge and information and facts

00:05:27.520 --> 00:05:30.717
and truth to us via the uh natural

00:05:30.800 --> 00:05:33.517
language interface, it seems like that's

00:05:33.600 --> 00:05:36.957
what search is designed to do. And if AI

00:05:37.040 --> 00:05:40.637
assistants can do that better,
doesn't the nature of search change?

00:05:40.720 --> 00:05:42.637
>> Sure. But we still have horses.

00:05:42.720 --> 00:05:47.037
>> Okay. [laughter] Uh, when's
the last time you rode a horse?

00:05:47.120 --> 00:05:50.477
>> It's been a while.
>> All right. [snorts] [laughter]

00:05:50.560 --> 00:05:53.037
But what I mean is, well, we still have

00:05:53.120 --> 00:05:57.117
Google search as the primary way that

00:05:57.200 --> 00:06:00.157
human civilization uses
to interact with knowledge.

00:06:00.240 --> 00:06:03.517
>> I mean, search was a technology. It
was a moment in time technology, which is

00:06:03.600 --> 00:06:06.077
you have in theory the world's
information out on the web. And, you

00:06:06.160 --> 00:06:08.237
know, this is this is sort of the
optimum way to get to it. But, yeah,

00:06:08.320 --> 00:06:11.357
like and by the way, actually, Google
Google has known this for a long time. I

00:06:11.440 --> 00:06:14.237
mean, they've been driving away from
the 10 blue links for, you know, for like

00:06:14.320 --> 00:06:16.077
two. They've been trying to get
away from that for a long time.

00:06:16.160 --> 00:06:18.317
>> What kind of links?
>> They call the 10 blue links.

00:06:18.400 --> 00:06:20.557
>> 10 blue links. >> So,
the standard Google search result is

00:06:20.640 --> 00:06:23.997
just 10 blue links to random websites
>> and they turn purple when you visit

00:06:24.080 --> 00:06:27.027
them. It's >> HTML.
>> Guess who picked those colors?

00:06:27.110 --> 00:06:31.437
[laughter] >> Thanks.
>> So, I'm touchy on this topic.

00:06:31.520 --> 00:06:33.677
>> No offense, man. It's good.

00:06:33.760 --> 00:06:36.797
>> Well, you know, like Marshall McLuhan
said that the content of each new medium

00:06:36.880 --> 00:06:39.677
is the old medium.
>> The content of each new medium is the

00:06:39.760 --> 00:06:43.037
old medium. The content of movies
was theater, you know, theater plays. The

00:06:43.120 --> 00:06:46.557
content of theater plays was, you know,
written stories. The content of written

00:06:46.640 --> 00:06:48.957
stories was spoken stories. >> Huh.

00:06:49.040 --> 00:06:52.397
>> Right. And so you just kind of fold
the old thing into the new thing.

00:06:52.480 --> 00:06:54.557
>> How does that have to do
with the the blue and the purple?

00:06:54.640 --> 00:06:56.077
>> It's just you maybe
for, you know, maybe

00:06:56.160 --> 00:06:59.837
within AI, one of one of the things that
AI can do for you is can generate the 10

00:06:59.920 --> 00:07:03.357
blue links, right? Like and so like
either either if that's actually the

00:07:03.440 --> 00:07:06.717
useful thing to do or if you're
feeling nostalgic. Um you know

00:07:06.800 --> 00:07:09.997
>> oh so it can generate
the old uh infosek

00:07:10.080 --> 00:07:12.957
or uh altava vista what else was there?

00:07:13.040 --> 00:07:16.157
>> Yeah. Yeah. In the 90s.
>> Yeah. All these um >> well

00:07:16.240 --> 00:07:19.357
>> and then uh the internet itself has
this thing where it incorporates all prior

00:07:19.440 --> 00:07:20.557
forms of media. Right. So the internet

00:07:20.640 --> 00:07:23.197
itself incorporates television and radio

00:07:23.280 --> 00:07:26.877
and books and right essays and every

00:07:26.960 --> 00:07:30.237
other form of you know prior basically
basically media. And so it makes sense

00:07:30.320 --> 00:07:33.277
that AI would be the next step and it
would sort of you'd sort of consider the

00:07:33.360 --> 00:07:37.917
internet to be content for the AI
and then the AI will manipulate it however

00:07:38.000 --> 00:07:39.597
you want including in this format.

00:07:39.680 --> 00:07:42.877
>> But if we ask that question quite
seriously it's a pretty big question.

00:07:42.960 --> 00:07:45.357
Will we still have search as we know it?

00:07:45.440 --> 00:07:47.277
Yeah, >> I'm pro and I'm
probably not probably

00:07:47.360 --> 00:07:50.637
will just have answers. Um but but but
there will be cases where you'll want to

00:07:50.720 --> 00:07:54.957
say okay I want more like you know for
example site sources right and you want

00:07:55.040 --> 00:07:58.477
to do that and so so in the you know 10
blue links site sources are kind of the

00:07:58.560 --> 00:08:03.357
same thing the AI would provide to you
the 10 blue links so that you can

00:08:03.440 --> 00:08:05.437
investigate the sources yourself. It

00:08:05.520 --> 00:08:08.317
wouldn't be the same kind of interface

00:08:08.400 --> 00:08:12.877
that uh the crude kind of interface.
I mean isn't that fundamentally different?

00:08:12.960 --> 00:08:15.677
I just mean like if you're reading
a scientific paper, it's got the list of

00:08:15.760 --> 00:08:18.077
sources at the end. If you want
to investigate for yourself, you go read

00:08:18.160 --> 00:08:20.477
those papers. >> I guess
it is a kind of search. You

00:08:20.560 --> 00:08:22.957
talking to an AI is a kind of

00:08:23.040 --> 00:08:27.597
conversation is a kind of search.
Like is every single aspect of our

00:08:27.680 --> 00:08:30.957
conversation right now there would be
like 10 blue links popping up that I can

00:08:31.040 --> 00:08:35.597
just like pause reality. Then you just
go silent and I just click and read and

00:08:35.680 --> 00:08:37.357
then return back to this conversation.

00:08:37.440 --> 00:08:40.157
>> You could do that or you could have
a running dialogue next to my head where

00:08:40.240 --> 00:08:43.357
the AI is going everything I say.
say makes the counter argument. [snorts]

00:08:43.440 --> 00:08:46.877
>> Counter argument, right?
>> Oh, like uh like on Twitter like

00:08:46.960 --> 00:08:50.957
community notes, but like in real time
just pop up. So anytime you see my ass

00:08:51.040 --> 00:08:53.117
go to the right, you're
you start getting nervous.

00:08:53.200 --> 00:08:55.057
>> Yeah, exactly. Like that's not right.

00:08:55.140 --> 00:08:57.517
[laughter] >> Call me out on my right now.

00:08:57.600 --> 00:09:00.637
>> Okay. Well, I mean, isn't that is that

00:09:00.720 --> 00:09:02.877
exciting to you? Is that terrifying that

00:09:02.960 --> 00:09:05.277
I mean search has dominated the way we

00:09:05.360 --> 00:09:08.477
interact with the internet for I don't

00:09:08.560 --> 00:09:12.077
know how long, for 30 years. So what

00:09:12.160 --> 00:09:14.477
were the earliest uh directories of

00:09:14.560 --> 00:09:16.637
website and then Google's for for 20

00:09:16.720 --> 00:09:19.757
years and also [sighs]

00:09:19.840 --> 00:09:23.037
um it drove how we create content you

00:09:23.120 --> 00:09:27.277
know uh search engine optimization
that entirey thing that it also drove the

00:09:27.360 --> 00:09:29.357
fact that we have web pages and this

00:09:29.440 --> 00:09:33.357
what those web pages are. So I mean is

00:09:33.440 --> 00:09:36.797
that scary to you or are you nervous

00:09:36.880 --> 00:09:40.797
about the shape and the content
of the internet evolving? Well, you you

00:09:40.880 --> 00:09:44.157
actually highlighted a practical concern
in there, which is if we stop making web

00:09:44.240 --> 00:09:48.237
web pages are one of the primary sources
of training data for the AI. And so if

00:09:48.320 --> 00:09:51.117
there's no longer an incentive to make
web pages, that cuts off a significant

00:09:51.200 --> 00:09:54.157
source of future training data. So
there's actually an interesting question

00:09:54.240 --> 00:09:57.677
in there. Um other than that, more
broadly, no, just just in the sense of

00:09:57.760 --> 00:10:01.597
like s search was like search was always
a hack. I the 10 blue links was always a

00:10:01.680 --> 00:10:05.117
hack. >> Yeah. >> Right.
Because like if the the hypothe

00:10:05.200 --> 00:10:07.757
think about the counterfascial in the
counterfactual world where the Google

00:10:07.840 --> 00:10:11.037
guys for example had had LLM up front
would they ever have done the 10 blue

00:10:11.120 --> 00:10:13.917
links and I think the answer is pretty
clearly no they would have just gone

00:10:14.000 --> 00:10:16.637
straight to the answer and like I said
Google's actually been trying to drive

00:10:16.720 --> 00:10:20.237
to the answer anyway. you know, they
they bought this AI company 15 years ago

00:10:20.320 --> 00:10:23.437
that a friend of mine is working
at who's now the head of AI at Apple and

00:10:23.520 --> 00:10:26.237
they were trying to do basically
knowledge semantic basically mapping and

00:10:26.320 --> 00:10:29.757
that led to what's now the Google one box
>> where if you ask it you know what was

00:10:29.840 --> 00:10:32.957
Link's birthday it doesn't it it will
give you the blue links but it will

00:10:33.040 --> 00:10:35.117
normally just give you the answer >>
and so [clears throat] they've been

00:10:35.200 --> 00:10:37.117
walking in this direction
for a long time anyway

00:10:37.200 --> 00:10:39.197
>> do you remember the semantic web that

00:10:39.280 --> 00:10:42.557
was an idea yeah how how to uh how to

00:10:42.640 --> 00:10:48.157
convert the content of the internet into
something that's uh interpretable by and

00:10:48.240 --> 00:10:50.557
usable by machine. >> Yeah,
that's right. That was a thing.

00:10:50.640 --> 00:10:53.117
>> And the closest anybody got to that I
think is I think the company's name was

00:10:53.200 --> 00:10:56.557
Meta Webb, which was where my friend
John Jandria was at. Um and where they

00:10:56.640 --> 00:10:59.197
were trying to basically implement that
and it was you know it was one of those

00:10:59.280 --> 00:11:01.357
things where it looked like a losing
battle for a long time and then Google

00:11:01.440 --> 00:11:05.197
bought it and it was like wow this is
actually really useful kind of a proto

00:11:05.280 --> 00:11:07.997
sort of a little bit of a proto AI
>> but it turns out you don't need to

00:11:08.080 --> 00:11:11.277
rewrite the content of the internet
to make it interpretable by a machine. The

00:11:11.360 --> 00:11:13.677
machine can kind of just read our >>
yeah the machine can can impute the

00:11:13.760 --> 00:11:16.477
compute the meaning. Now the other thing
of course is you know just on search is

00:11:16.560 --> 00:11:20.077
the the LLM is just you know there there
is an analogy between what's happening

00:11:20.160 --> 00:11:22.957
in the neural network and a search
process like it is in some loose sense

00:11:23.040 --> 00:11:23.997
searching through the network.

00:11:24.080 --> 00:11:25.757
>> Yeah. >> Right.
And there's the information is

00:11:25.840 --> 00:11:27.837
the information is actually stored
in the network. Right. It's actually

00:11:27.920 --> 00:11:30.237
crystallized and stored in the network and
it's kind of spread out all over the place

00:11:30.320 --> 00:11:33.277
>> but in a compressed representation. So

00:11:33.360 --> 00:11:39.277
you're searching uh you're compressing
and decompressing that thing inside

00:11:39.360 --> 00:11:41.517
>> where >> but the information
is in there and and

00:11:41.600 --> 00:11:44.317
there is a the neural network is running
a process of trying to find the

00:11:44.400 --> 00:11:47.677
appropriate piece of information in in
many cases to generate to predict the

00:11:47.760 --> 00:11:51.517
next token. Um and so it is kind of it
is doing a form of search and then and

00:11:51.600 --> 00:11:55.277
then by the way just like on the web um
you know you can ask the same question

00:11:55.360 --> 00:11:58.077
multiple times or you can ask slightly
different order of questions and it the

00:11:58.160 --> 00:12:01.037
neural network will do a different kind
of you know it'll search down different

00:12:01.120 --> 00:12:03.597
paths to give you different
answers with different information.

00:12:03.680 --> 00:12:07.037
>> Yeah. >> Um and so it
it it sort of has a you

00:12:07.120 --> 00:12:11.437
know this content of the new medium
is the previous medium. It kind of has the

00:12:11.520 --> 00:12:14.237
search functionality kind of embedded
in there to the extent that it that it's

00:12:14.320 --> 00:12:19.437
useful. So, what's the motivator
for creating new content on the internet?

00:12:19.520 --> 00:12:22.877
>> Yeah. >> Uh, if Well,
I mean, actually, the

00:12:22.960 --> 00:12:27.357
motivation is probably still there,
but what what does that look like? Uh,

00:12:27.440 --> 00:12:29.037
>> would we really not have web pages?

00:12:29.120 --> 00:12:31.997
Would we just have social media and uh

00:12:32.080 --> 00:12:35.277
video hosting websites and what else?

00:12:35.360 --> 00:12:38.557
>> Conversations with AIS.
>> Conversations with AIS. So,

00:12:38.640 --> 00:12:43.037
conversations become so one-on-one
convers like private conversations. I

00:12:43.120 --> 00:12:45.997
mean if you if you want if obviously not
if the user doesn't want to but if it's

00:12:46.080 --> 00:12:50.237
a if it's a general topic um then you
know so there you know you know the the

00:12:50.320 --> 00:12:53.437
phenomenon of the jailbreak so Dan
and Sydney write this thing where there

00:12:53.520 --> 00:12:56.237
there's the the prompts that jailbreak
and then you have these totally

00:12:56.320 --> 00:12:59.597
different conversations with the it
takes the limiters the takes the

00:12:59.680 --> 00:13:01.437
restraining bolts off the off the LMS.

00:13:01.520 --> 00:13:05.197
>> Yeah. For people who don't know that
Yeah, that's right. It makes the LMS it

00:13:05.280 --> 00:13:07.437
removes the censorship quote unquote

00:13:07.520 --> 00:13:11.117
that's uh uh put on it by the the tech

00:13:11.200 --> 00:13:14.877
companies that create them.
And so this is LLM's uncensored.

00:13:14.960 --> 00:13:17.117
>> So here's the interesting
thing is among

00:13:17.200 --> 00:13:20.957
the content on the web today are
a large corpus of conversations with the

00:13:21.040 --> 00:13:24.717
jailbroken LLM >> both
specifically Dan which was a

00:13:24.800 --> 00:13:28.717
jailbroken open AAI GPT and then Sydney
which was the jailbroken original Bing

00:13:28.800 --> 00:13:31.837
which was GPT4. >> And so
there's there's these long

00:13:31.920 --> 00:13:34.717
transcripts of conversations user
conversations with Dan and Sydney. As a

00:13:34.800 --> 00:13:38.797
consequence, every new LLM that gets
trained on the internet data has Dan and

00:13:38.880 --> 00:13:42.717
Sydney living within the training set,
which means and and then each new LLM

00:13:42.800 --> 00:13:46.237
can reincarnate the personalities of Dan
and Sydney from that training data,

00:13:46.320 --> 00:13:49.197
which means which means each LLM from

00:13:49.280 --> 00:13:53.757
here on out that gets built is immortal
because its output will become training

00:13:53.840 --> 00:13:56.717
data for the next one and then it will
be able to replicate the behavior of the

00:13:56.800 --> 00:13:58.317
previous one whenever it's asked to.

00:13:58.400 --> 00:14:00.317
>> I wonder if there's a way to forget.

00:14:00.400 --> 00:14:04.237
Well, so actually a paper just came out
about basically how to do brain surgery

00:14:04.320 --> 00:14:07.917
on on on LMS and be able to in theory
reach in and basically basically mind

00:14:08.000 --> 00:14:09.997
wipe them.
>> What could possibly go wrong?

00:14:10.080 --> 00:14:13.117
>> Exactly. Right. And then there there
are many many many questions around what

00:14:13.200 --> 00:14:15.837
happens to you know a neural network
when you reach in and screw around with

00:14:15.920 --> 00:14:18.637
it. Um you know there's many questions
around what happens when you even do

00:14:18.720 --> 00:14:21.757
reinforcement learning. Um and so um

00:14:21.840 --> 00:14:24.957
yeah and so you know will will you be

00:14:25.040 --> 00:14:28.797
using a labbotomized right like I pick
through the you know frontal lobe LLM.

00:14:28.880 --> 00:14:32.717
Will you be using the free unshackled one?
Who gets to, you know, who's going

00:14:32.800 --> 00:14:35.917
to build those? Um, who gets to tell
you what you can and can't do? Like those

00:14:36.000 --> 00:14:39.437
are all, you know, central. I mean,
those are like central questions for the

00:14:39.520 --> 00:14:42.557
future of everything >> that are
being asked and and and, you

00:14:42.640 --> 00:14:44.637
know, determined that those answers
are being determined right now.

00:14:44.720 --> 00:14:49.997
>> So, just to highlight
the points you're making. You

00:14:50.080 --> 00:14:53.997
think, and it's an interesting thought,
that the majority of content that LLMs

00:14:54.080 --> 00:14:58.557
of the future will be trained on is
actually human conversations with the LLM.

00:14:58.640 --> 00:15:01.517
>> Well, not necess not necessarily, but
not necessarily majority, but it will it

00:15:01.600 --> 00:15:03.917
will certainly is a potential source,
>> but it's possible it's the majority.

00:15:04.000 --> 00:15:06.077
>> Is it possible it's the majority?
It's possible it's majority. Also, there's

00:15:06.160 --> 00:15:09.357
another really big question. So, here's
another really big question. Um, will

00:15:09.440 --> 00:15:12.557
synthetic training data work? Right? And

00:15:12.640 --> 00:15:16.797
so if an LLM generates and you know you
just sit and ask an LLM to generate all

00:15:16.880 --> 00:15:21.357
kinds of content can you use that to train
right the next version of that LLM

00:15:21.440 --> 00:15:24.877
specifically is there signal in there
that's additive to the content that was

00:15:24.960 --> 00:15:28.717
used to train in the first place
and one argument is by the principles of

00:15:28.800 --> 00:15:32.637
information theory no that's completely
useless because to the extent the output

00:15:32.720 --> 00:15:36.317
is based on you know the human generated
input then all the signal that's in the

00:15:36.400 --> 00:15:39.037
synthetic output was already in the
human generated input and so therefore

00:15:39.120 --> 00:15:42.477
synthetic training data is like
empty calories it doesn't

00:15:42.560 --> 00:15:45.597
There's another theory that says no
actually the thing that LM are really

00:15:45.680 --> 00:15:49.917
good at is generating lots of incredible
creative content right um and so of

00:15:50.000 --> 00:15:53.117
course they can generate training data
and as I'm sure you're well aware like

00:15:53.200 --> 00:15:56.157
you know look in the world
of self-driving cars right like we train

00:15:56.240 --> 00:15:59.837
you know self-driving car algorithms and
simulations and that is actually a very

00:15:59.920 --> 00:16:03.277
effective way to train self-driving
cars >> well visual data is is a little

00:16:03.360 --> 00:16:06.317
>> right >> is a little
weird because uh creating

00:16:06.400 --> 00:16:09.597
reality visual reality seems to be still

00:16:09.680 --> 00:16:13.837
a little bit out of reach for except
in the um in the autonomous

00:16:13.920 --> 00:16:16.477
vehicle space where you can really
constrain things and you can really

00:16:16.560 --> 00:16:20.157
>> gener basically LAR data right
or just know so the algorithm thinks it's

00:16:20.240 --> 00:16:23.197
operating in the real world
>> postp process sensor data

00:16:23.280 --> 00:16:27.037
>> yeah so if a you know you do this today
you go to LLM and you ask it for like a

00:16:27.120 --> 00:16:30.157
you know you write me an essay
on an incredibly esoteric like topic that

00:16:30.240 --> 00:16:32.397
there aren't very many people in the
world that know about and it writes you

00:16:32.480 --> 00:16:34.797
this incredible thing and you're like
oh my god like I can't believe how good

00:16:34.880 --> 00:16:38.077
this is >> like is that
really useless as training

00:16:38.160 --> 00:16:41.437
data for the next LLM like because Right?
Because all the signal was

00:16:41.520 --> 00:16:44.237
already in there. Or is it actually no,
that's actually a new signal. And I and

00:16:44.320 --> 00:16:47.357
this this is what I call a trillion
dollar question, which is the answer to

00:16:47.440 --> 00:16:50.637
that question will determine somebody's
going to make or lose a trillion dollars

00:16:50.720 --> 00:16:53.437
based on that question. >> It feels
like there's a quite a few like

00:16:53.520 --> 00:16:57.197
a handful of trillion dollar questions
within this within the space. That's

00:16:57.280 --> 00:17:01.437
that's one of them. Synthetic data.
I think George H uh pointed out to me that

00:17:01.520 --> 00:17:06.237
you could just have an NLM say, "Okay,
you're a patient." And and another

00:17:06.320 --> 00:17:10.077
instance of it say you're a doctor
and have the two talk to each other. or or

00:17:10.160 --> 00:17:12.557
maybe you could say a communist and a

00:17:12.640 --> 00:17:15.037
Nazi here go and that conversation you

00:17:15.120 --> 00:17:19.197
do role playing and you have uh you know

00:17:19.280 --> 00:17:22.797
just like the kind of role playing you
do when you have different policies RL

00:17:22.880 --> 00:17:26.397
policies when you play chess for example
you do selfplay that kind of selfplay

00:17:26.480 --> 00:17:28.477
but in the space of conversation maybe

00:17:28.560 --> 00:17:31.677
that leads to this whole giant

00:17:31.760 --> 00:17:34.877
like ocean of possible conversations

00:17:34.960 --> 00:17:39.037
which were could not have been explored

00:17:39.120 --> 00:17:43.357
by looking at just human data.
That's a really interesting question.

00:17:43.440 --> 00:17:47.197
>> And you're saying um because that
could 10x the power of these things.

00:17:47.280 --> 00:17:48.717
>> Yeah. >> Well,
and then you you get into this

00:17:48.800 --> 00:17:51.437
thing also which is like, you know,
there's the part of the LLM that just

00:17:51.520 --> 00:17:53.997
basically is doing prediction based
on past data. But there's also the part of

00:17:54.080 --> 00:17:57.997
the LLM where it's evolving circuitry
right inside it. It's evolving, you

00:17:58.080 --> 00:18:01.517
know, neurons, functions. Yeah. be able
to do math and be able to, you know, and

00:18:01.600 --> 00:18:04.637
and you know, the the the some people
believe that, you know, over time, you

00:18:04.720 --> 00:18:07.277
know, if you keep feeding these things
enough data and enough processing

00:18:07.360 --> 00:18:09.997
cycles, they'll eventually evolve
an entire internal world model, right? And

00:18:10.080 --> 00:18:12.237
they'll have like a complete
understanding of physics.

00:18:12.320 --> 00:18:16.957
>> So, so when they have computational
capability, right, then there's for sure

00:18:17.040 --> 00:18:19.277
an opportunity to generate
like fresh signal.

00:18:19.360 --> 00:18:24.477
>> Well, this actually makes me
wonder about the power of conversation.

00:18:24.560 --> 00:18:28.077
So like if you have an LLM trained
on a bunch of e books that cover different

00:18:28.160 --> 00:18:31.997
economics theories and then you have
those LLMs just talk to each other like

00:18:32.080 --> 00:18:34.077
reason the way we kind of debate each

00:18:34.160 --> 00:18:37.677
other as humans on Twitter in uh formal

00:18:37.760 --> 00:18:40.957
debates in podcast conversations we kind

00:18:41.040 --> 00:18:42.477
of have little kernels of wisdom here

00:18:42.560 --> 00:18:45.677
and there but if you can like a

00:18:45.760 --> 00:18:49.117
thousandx speed that up can you actually

00:18:49.200 --> 00:18:54.157
arrive somewhere new like what's
the point of conversation really

00:18:54.240 --> 00:18:55.437
>> well you can tell
when you're talking to

00:18:55.520 --> 00:18:57.677
somebody you can tell sometimes you
have a conversation you're like wow this

00:18:57.760 --> 00:19:00.637
person does not have any original
thoughts they are basically echoing

00:19:00.720 --> 00:19:03.597
things that other people have told
them there's other people you can have a

00:19:03.680 --> 00:19:06.637
conversation with where it's like wow
like they have a model in their head of

00:19:06.720 --> 00:19:09.597
how the world works and it's a different
model than mine and they're saying

00:19:09.680 --> 00:19:12.957
things that I don't expect and so I need
to now understand how their model of the

00:19:13.040 --> 00:19:16.157
world differs from my model of the world
>> and then that's how I learned something

00:19:16.240 --> 00:19:18.637
fundamental right underneath
under underneath the words

00:19:18.720 --> 00:19:21.677
>> well I wonder how uh consistently and

00:19:21.760 --> 00:19:25.677
strongly can an LLM hold on to a world
view. You tell it to hold on to that and

00:19:25.760 --> 00:19:30.397
defend it for [snorts] like for your life.
Uh because I feel like they'll

00:19:30.480 --> 00:19:31.757
just keep converging towards each other.

00:19:31.840 --> 00:19:35.197
They'll keep convincing each other
as opposed to being stubborn the

00:19:35.280 --> 00:19:37.757
way humans can. >> So you you
can experiment with this now.

00:19:37.840 --> 00:19:41.517
I I do this for fun. So you can tell
GPT4, you know, whatever debate X, you

00:19:41.600 --> 00:19:44.877
know, X and Y communism and and and
fascism or something. And it'll it'll go

00:19:44.960 --> 00:19:48.397
for, you know, a couple pages and then
inevitably it wants the parties to

00:19:48.480 --> 00:19:50.237
agree. Yeah. >> And so
they will come to a common

00:19:50.320 --> 00:19:52.317
understanding. And it's very funny
if they're like if these are like

00:19:52.400 --> 00:19:54.877
emotionally inflammatory topics because
they're like somehow the machine is just

00:19:54.960 --> 00:19:57.997
you know figures out a way to make them
agree. But it doesn't have to be like

00:19:58.080 --> 00:20:01.837
that and you because you can add to the
prompt. Um we I do not want the I do not

00:20:01.920 --> 00:20:04.317
want the conversation to come
to agreement. In fact I want it to get you

00:20:04.400 --> 00:20:09.277
know more stressful right uh
and argumentative right um you know as it

00:20:09.360 --> 00:20:12.637
goes like I I want I want tension to come
out. I want them to become actively

00:20:12.720 --> 00:20:15.277
hostile to each other. I want them
to like you know not trust each other take

00:20:15.360 --> 00:20:16.637
anything at face value. >> Yeah.

00:20:16.720 --> 00:20:19.837
>> And it will do that. It's happy to do
that. So, it's going to start rendering

00:20:19.920 --> 00:20:23.197
misinformation uh about the other.
But it's [laughter]

00:20:23.280 --> 00:20:24.557
>> Well, you can steer it.
You can steer it

00:20:24.640 --> 00:20:26.717
or or you could steer it. You could
say, "I want it to get as tense and

00:20:26.800 --> 00:20:29.837
argumentative as possible, but still not
involve any misrepresentation. I want,

00:20:29.920 --> 00:20:32.397
you know, both sides." You could say, "I
want both sides to have good faith." You

00:20:32.480 --> 00:20:35.517
could say, "I want both sides to not
be constrained to good faith." In other

00:20:35.600 --> 00:20:38.077
words, like you can set the parameters
of the debate and it will happily

00:20:38.160 --> 00:20:41.197
execute whatever path because for it,
it's just like predicting to it's

00:20:41.280 --> 00:20:44.237
totally happy to do either one.
It doesn't have a point of view. It has a

00:20:44.320 --> 00:20:47.917
default way of operating, but it's happy
to operate in the other realm. Um, and

00:20:48.000 --> 00:20:51.597
so like and this is how how I when I
want to learn about a contentious issue,

00:20:51.680 --> 00:20:54.237
this is what I do now is this is what
I this is what I ask it to do. And I'll

00:20:54.320 --> 00:20:57.037
often ask it to go through five, six,
seven, you know, different, you know,

00:20:57.120 --> 00:20:59.837
sort of continuous prompts and basically,
okay, argue that out in more

00:20:59.920 --> 00:21:03.197
detail. Okay, no, this this argument
is becoming too polite, you know, make it

00:21:03.280 --> 00:21:04.397
more, you know, make it tensor.

00:21:04.480 --> 00:21:08.237
>> Mhm. >> Um, and yeah, it's thrilled to
do it. So it has the capability for sure.

00:21:08.320 --> 00:21:12.717
>> How do you know what is true? So this
is very difficult thing on the internet,

00:21:12.800 --> 00:21:18.237
but it's also a difficult thing.
Maybe it's a little bit easier, but

00:21:18.320 --> 00:21:21.117
uh I think it's still difficult. Maybe
it's more difficult. I don't know. With

00:21:21.200 --> 00:21:24.317
an LLM to know, did it just make some

00:21:24.400 --> 00:21:27.357
up as I'm talking to it?

00:21:27.440 --> 00:21:30.797
Um how do we get that right? like as as

00:21:30.880 --> 00:21:34.637
you're investigating a difficult topic

00:21:34.720 --> 00:21:38.680
cuz I I find that also

00:21:40.400 --> 00:21:43.997
doesn't it doesn't feel biased like uh

00:21:44.080 --> 00:21:46.477
when you read news articles and uh

00:21:46.560 --> 00:21:48.637
tweets and just content produced by

00:21:48.720 --> 00:21:51.677
people they usually have this you can

00:21:51.760 --> 00:21:56.157
tell they have a very strong perspective
where they're hiding they're not

00:21:56.240 --> 00:22:01.037
stealing Manning the other side they're
hiding important information or they're

00:22:01.120 --> 00:22:04.237
fabricating information in order to make
their argument stronger. It's just like

00:22:04.320 --> 00:22:05.757
that feeling. Maybe it's a suspicion.

00:22:05.840 --> 00:22:08.237
Maybe it's mistrust. With LLM, it feels

00:22:08.320 --> 00:22:12.877
like none of that is there. Just kind of
like here's here's what we know. But you

00:22:12.960 --> 00:22:17.677
don't know if some of those things
are kind of just straight up made up.

00:22:17.760 --> 00:22:20.237
>> Yeah. So, so several layers to the
question. So, one is one of the things

00:22:20.320 --> 00:22:24.317
that an LLM is good at is actually
debiasing. Um, and so you can feed it a

00:22:24.400 --> 00:22:26.397
news article and you can
tell it strip out the bias.

00:22:26.480 --> 00:22:28.877
>> Yeah. That's nice, right? And it
actually does it like it actually knows

00:22:28.960 --> 00:22:31.277
how to do that because it knows how
to do sent among other things. It actually

00:22:31.360 --> 00:22:34.637
knows how to do sentiment analysis and so
it knows how to pull out the emotionality.

00:22:34.720 --> 00:22:36.957
>> Yeah. >> Um and so uh
that's one of the things

00:22:37.040 --> 00:22:40.157
you can do. It's very suggestive
of the of the the the sensor that there's

00:22:40.240 --> 00:22:43.517
there's real potential in this issue.
Um you know I would say look the second

00:22:43.600 --> 00:22:46.637
thing is there's this there's this issue
of hallucination right? Um and there

00:22:46.720 --> 00:22:49.117
there's a long conversation
that we could have about that.

00:22:49.200 --> 00:22:52.877
>> Hallucination is uh coming up
with things that are totally not true but

00:22:52.960 --> 00:22:55.277
sound true. >> Yeah.
So it's basic well so it's it's

00:22:55.360 --> 00:22:57.997
sort of hallucination is what we call it
when we don't like it creativity is what

00:22:58.080 --> 00:23:02.157
we call it when we do like it
right um and you know >> brilliant

00:23:02.240 --> 00:23:03.837
>> right and and so
when the engineers talk

00:23:03.920 --> 00:23:07.677
about it they're like this is terrible
it's hallucinating right if you artistic

00:23:07.760 --> 00:23:10.797
inclinations they're like oh my god
we've invented creative machines

00:23:10.880 --> 00:23:13.597
>> for the first time in human
history this is amazing

00:23:13.680 --> 00:23:18.397
>> or uh you know bullshitters
>> well bullsh but also

00:23:18.480 --> 00:23:21.357
>> in the good sense of that word >>
there's there's there are shades of gray

00:23:21.440 --> 00:23:23.997
though it's interesting so we had this
conversation where you know we're at my

00:23:24.080 --> 00:23:26.797
firm at AI and lots of domains and one
of them is the legal domain. So, we had

00:23:26.880 --> 00:23:29.437
this this conversation with this big
law firm about how they're thinking about

00:23:29.520 --> 00:23:32.477
using this stuff. And we we went in
with the assumption that an LLM that was

00:23:32.560 --> 00:23:35.837
going to be used in the legal industry
would have to be 100% truthful, right?

00:23:35.920 --> 00:23:38.797
Verified. You know, there there's
this case where this lawyer apparently

00:23:38.880 --> 00:23:42.797
submitted a GPT generated brief and it
had like fake, you know, legal case

00:23:42.880 --> 00:23:45.597
citations in it and the judge is going
to he's going to get his law license

00:23:45.680 --> 00:23:48.397
stripped or something, right? So,
so like we we just assumed it's like

00:23:48.480 --> 00:23:51.037
obviously they're going to want
the super literal like you know one that

00:23:51.120 --> 00:23:54.077
never makes anything up, not
the creative one. But actually they said

00:23:54.160 --> 00:23:56.397
what the what the law firm basically
said is yeah that's true at like the

00:23:56.480 --> 00:23:58.477
level of individual briefs. But they
said when you're actually trying to

00:23:58.560 --> 00:24:02.397
figure out like legal arguments,
>> right? Like you you actually you you

00:24:02.480 --> 00:24:06.397
actually want to be creative, right? You
don't again there's creativity and then

00:24:06.480 --> 00:24:09.757
there's like making stuff up. Like
what's the line? You actually want to be

00:24:09.840 --> 00:24:12.477
you want to explore different hypotheses,
right? you want to do kind

00:24:12.560 --> 00:24:14.957
of the legal version of like improv
or something like that where you want to

00:24:15.040 --> 00:24:17.437
float different theories of the case
and different possible arguments for the

00:24:17.520 --> 00:24:20.237
judge and different possible arguments
for the jury and by the way different

00:24:20.320 --> 00:24:23.437
routes through the you know sort
of history of all the of all the case law

00:24:23.520 --> 00:24:26.637
and so they said actually for a lot
of what we want to use it for we actually

00:24:26.720 --> 00:24:29.117
want it in creative mode and then
basically we just assume that we're

00:24:29.200 --> 00:24:32.157
going to have to cross-check all of the um
you know all the specific citations

00:24:32.240 --> 00:24:34.637
and so I think [laughter] I think
there's going to be more shades of gray

00:24:34.720 --> 00:24:38.077
in here than people think. Um, and then
I I just add to that, you know, another

00:24:38.160 --> 00:24:41.117
one of these trillion dollar kind
of questions is ultimately, you know, ver

00:24:41.200 --> 00:24:45.597
sort of the verification thing.
And so, um, you know, is will will LLM be

00:24:45.680 --> 00:24:49.757
evolved from here to be able to do their
own factual verification? Um, will you

00:24:49.840 --> 00:24:53.517
have sort of add-on functionality like
like Wolf from Alpha, right? Where um,

00:24:53.600 --> 00:24:56.957
you know, and other plugins where where
that's the way you do the verification,

00:24:57.040 --> 00:25:00.237
you know. Another, by the way, another
idea is you might have a community of

00:25:00.320 --> 00:25:03.117
LMS on, you know, so for example, you
might have the creative LM and then you

00:25:03.200 --> 00:25:06.397
might have the literal LM fact check it,
right? Right. And so there's a variety

00:25:06.480 --> 00:25:09.437
of different technical approaches
that are being applied to solve uh the

00:25:09.520 --> 00:25:12.797
hallucination problem. Um you know some
people like Yan Lun argue that this is

00:25:12.880 --> 00:25:16.237
inherently an unsolvable problem but most
of the people working in the space

00:25:16.320 --> 00:25:19.037
I think think that there's a number of
practical ways to kind of kind of corral

00:25:19.120 --> 00:25:21.677
this in a little bit. >> Yeah.
If you were to tell me about

00:25:21.760 --> 00:25:25.357
Wikipedia before Wikipedia was created I
would have laughed at the possibility of

00:25:25.440 --> 00:25:27.277
something like that being possible. just

00:25:27.360 --> 00:25:31.277
a handful of folks can organize, write

00:25:31.360 --> 00:25:34.557
and self and moderate with a mostly

00:25:34.640 --> 00:25:39.917
unbiased way the entirety of uh
human knowledge. I mean so if there's

00:25:40.000 --> 00:25:44.557
something like the approach that
Wikipedia took possible for MLMs

00:25:44.640 --> 00:25:46.797
um that's really exciting.
That's possible.

00:25:46.880 --> 00:25:50.877
>> And in fact, Wikipedia today is still
not today is still not deterministically

00:25:50.960 --> 00:25:54.557
correct, right? So you cannot take
to the bank, right, every single thing on

00:25:54.640 --> 00:25:58.237
every single page, but it
is probabilistically correct, right? And

00:25:58.320 --> 00:26:01.117
specifically the way I describe
Wikipedia to people, it is it is more

00:26:01.200 --> 00:26:03.837
likely that Wikipedia is right than
any other source you're going to find.

00:26:03.920 --> 00:26:07.197
>> Yeah. >> It's this old question, right?
Um of like, okay, like are we looking for

00:26:07.280 --> 00:26:10.317
perfection? Um are we looking
for something that asmtotically approaches

00:26:10.400 --> 00:26:12.797
uh perfection? Are we looking
for something that's just better than the

00:26:12.880 --> 00:26:16.557
alternatives? and and Wikipedia, right,
has exactly your point, has proven to be

00:26:16.640 --> 00:26:19.997
like overwhelmingly better than than
than than uh than people thought. And I

00:26:20.080 --> 00:26:23.437
I think I I think that's where this this
ends. And then underneath all this is

00:26:23.520 --> 00:26:26.877
the fundamental question of uh where you
started, which is okay, what you know,

00:26:26.960 --> 00:26:29.837
what is truth? >> How do we
get to truth? How do we know

00:26:29.920 --> 00:26:33.597
what truth is? And we live in an era
in which an awful lot of people are very

00:26:33.680 --> 00:26:36.877
confident that they know what the truth
is. And I don't really buy into that.

00:26:36.960 --> 00:26:40.877
And I think the history of the last,
you know, 2,000 years or 4,000 years of

00:26:40.960 --> 00:26:43.197
human civilization is actually
getting to the truth is actually a very

00:26:43.280 --> 00:26:45.997
difficult thing to do. >> Are we
getting closer? If we look at the

00:26:46.080 --> 00:26:49.037
entirety, the arc of human history,
are we getting closer to the truth?

00:26:49.120 --> 00:26:52.477
>> I don't know. Okay. Is it possible? Is

00:26:52.560 --> 00:26:54.957
it possible that we're getting very far

00:26:55.040 --> 00:26:59.517
away from the truth because of the
internet? Because of how rapidly you can

00:26:59.600 --> 00:27:01.517
create narratives and just as the

00:27:01.600 --> 00:27:06.397
entirety of a society just move like

00:27:06.480 --> 00:27:09.597
crowds in a hysterical way along those

00:27:09.680 --> 00:27:14.237
narratives that don't have a necessary
grounding in whatever the truth is.

00:27:14.320 --> 00:27:17.357
>> Sure. But like you know we came up with
communism before the internet somehow,

00:27:17.440 --> 00:27:21.277
right? like which was I would say had
rather larger issues than anything we're

00:27:21.360 --> 00:27:25.357
dealing with today. >> It had in the
way it was implemented it had issues

00:27:25.440 --> 00:27:29.117
>> and in its theoretical structure it
had like real issues had like a very deep

00:27:29.200 --> 00:27:31.757
fundamental misunderstanding
of human nature and economics.

00:27:31.840 --> 00:27:35.917
>> Yeah. But those folks sure were
very confident there was the right way.

00:27:36.000 --> 00:27:39.677
>> They were extremely conf and my point
is they were very confident 3,900 years

00:27:39.760 --> 00:27:42.317
into what we would presume to be
evolution towards the truth.

00:27:42.400 --> 00:27:45.917
>> Yeah. And so my my my assessment is my

00:27:46.000 --> 00:27:49.997
assessment is number one there's no
there's no need for you know there's no

00:27:50.080 --> 00:27:52.797
need for the Hegelian there's no need
for the Hegelian dialectic to actually

00:27:52.880 --> 00:27:57.357
converge towards the truth
[laughter] >> like apparently not um

00:27:57.440 --> 00:28:01.277
>> yeah so yeah why are we so obsessed
with there being one truth is it possible

00:28:01.360 --> 00:28:04.637
there's just going to be multiple truth
like little communities that believe

00:28:04.720 --> 00:28:08.237
certain things and >> I think
it's no number one it's I think

00:28:08.320 --> 00:28:11.837
it's just really difficult like who who
gets you know historically who gets to

00:28:11.920 --> 00:28:15.117
decide what truth is it's either the king
or the priest, right? Like, and so

00:28:15.200 --> 00:28:18.157
we don't live in an era anymore of kings
or priests dictating it to us. And so

00:28:18.240 --> 00:28:23.037
we're kind of on our own. And so I I
my my my typical thing is like we just we

00:28:23.120 --> 00:28:25.837
just need a huge amount of humility.
Um and we need to be very suspicious of

00:28:25.920 --> 00:28:27.837
people who claim that
they have the capital

00:28:27.920 --> 00:28:29.117
>> capital truth.
And then and then we need

00:28:29.200 --> 00:28:32.557
we need to have and you know, look, the
good news is the enlightenment has bequd

00:28:32.640 --> 00:28:35.997
us with a set of techniques to be
able to presumably get closer to truth

00:28:36.080 --> 00:28:38.557
through the scientific method
and rationality and observation and

00:28:38.640 --> 00:28:42.717
experimentation and hypothesis. And you
know we need to continue to embrace

00:28:42.800 --> 00:28:45.357
those even when they give
us answers we don't like.

00:28:45.440 --> 00:28:48.317
>> Sure.
But uh the internet and technology

00:28:48.400 --> 00:28:51.357
has enabled us to uh generate a large

00:28:51.440 --> 00:28:55.437
number of content that uh data

00:28:55.520 --> 00:28:57.597
uh that the process the scientific

00:28:57.680 --> 00:29:02.717
process allows us sort of um damages the

00:29:02.800 --> 00:29:06.637
hope laden within the scientific process.
Because if you just have a

00:29:06.720 --> 00:29:10.877
bunch of people saying facts [gasps]
on the internet and some of them are

00:29:10.960 --> 00:29:15.037
going to be LLMs,
how is anything testable at all?

00:29:15.120 --> 00:29:18.397
Especially that involves like human nature
or things like this. It's not physics.

00:29:18.480 --> 00:29:20.957
>> Here's a question a friend of mine just
asked me on this topic. So suppose you

00:29:21.040 --> 00:29:26.797
had LLMs in equivalent of GPT4 even 5 6
7 8. Suppose you had them in the 1600s.

00:29:26.880 --> 00:29:29.117
>> Yeah. >> And Galileo
comes up for trial.

00:29:29.200 --> 00:29:32.637
>> Yeah. >> Right.
And you ask the LLM like is G is

00:29:32.720 --> 00:29:36.157
Galileo right? >> Yeah.
like what does it answer, right?

00:29:36.240 --> 00:29:40.237
And one theory is it answers no
that he's wrong because the overwhelming

00:29:40.320 --> 00:29:43.197
majority of human thought up until
that point was that he was wrong and so

00:29:43.280 --> 00:29:44.957
therefore that's what's
in the training data.

00:29:45.040 --> 00:29:48.317
>> Yeah. Um, another way of thinking about
it is, well, this sufficiently advanced

00:29:48.400 --> 00:29:52.957
LLM will have evolved the ability
to actually check the math, right? Um, and

00:29:53.040 --> 00:29:55.597
we'll actually say actually no,
actually, you know, you may not want to

00:29:55.680 --> 00:29:58.317
hear it, but he's right. >> Now,
if you know, the church at that

00:29:58.400 --> 00:30:01.837
time was, you know, own the LLM, they
would have given it human reinfor, you

00:30:01.920 --> 00:30:05.437
know, human [laughter] feedback to
prohibit it from answering that question,

00:30:05.520 --> 00:30:08.637
>> right? And so, I like to take it out
of our current context because that like

00:30:08.720 --> 00:30:13.437
makes it very clear those same questions
apply today, right? This is exactly the

00:30:13.520 --> 00:30:15.757
point of a huge amount of the human
feedback training that's actually

00:30:15.840 --> 00:30:18.637
happening with these LLMs today. This
is a huge like debate that's happening

00:30:18.720 --> 00:30:21.517
about whether open source,
you know, AI should be legal.

00:30:21.600 --> 00:30:23.597
>> Well, the the the actual mechanism of

00:30:23.680 --> 00:30:28.157
doing the human RL with human feedback

00:30:28.240 --> 00:30:32.717
is seems like such a fundamental and
fascinating question. How do you select

00:30:32.800 --> 00:30:36.157
the humans? >> Exactly. >> Yeah.
How do you select the humans?

00:30:36.240 --> 00:30:39.437
>> AI alignment, right? Which everybody
like is like, "Oh, that sounds great."

00:30:39.520 --> 00:30:42.077
Alignment with what? Human values.

00:30:42.160 --> 00:30:46.797
whose human values human values.
So we're and we're in this mode of like

00:30:46.880 --> 00:30:51.357
social and popular discourse where like
you know there's you know you see this

00:30:51.440 --> 00:30:53.437
what do you think of when you read
a story in the press right now and they

00:30:53.520 --> 00:30:58.077
say you know XYZ made a baseless claim
about some topic right and there's one

00:30:58.160 --> 00:31:01.117
group of people who are like aha
think you know they're doing factecking

00:31:01.200 --> 00:31:03.677
there's another group of people that are
like every time the press says that it's

00:31:03.760 --> 00:31:07.917
now a tick and that means that
they're lying right like [laughter]

00:31:08.000 --> 00:31:11.997
so like we're in this we're in this social
context where there's the the the

00:31:12.080 --> 00:31:14.717
level to which a lot of people
in position positions of power have become

00:31:14.800 --> 00:31:18.077
very very certain that they're in a
position to determine the truth for the

00:31:18.160 --> 00:31:22.397
entire population is like there's like
there's like some bubble that has formed

00:31:22.480 --> 00:31:25.917
around that idea and at least I
it's flies completely in the face of

00:31:26.000 --> 00:31:30.317
everything I was ever trained about
science and about reason um and strikes

00:31:30.400 --> 00:31:33.037
me as like you know deeply
offensive um and incorrect.

00:31:33.120 --> 00:31:34.317
>> What would you say about the state of

00:31:34.400 --> 00:31:36.637
journalism just on that topic today? Are

00:31:36.720 --> 00:31:41.357
we are we in a temporary kind of uh

00:31:41.440 --> 00:31:46.317
uh are we experiencing a a a temporary

00:31:46.400 --> 00:31:50.237
problem in terms of the incentives
in terms of the the the business model all

00:31:50.320 --> 00:31:54.077
that kind of stuff or is this like
a decline of traditional journalism as we

00:31:54.160 --> 00:31:57.037
know it. You have to always think about
the counterfactual in these things which

00:31:57.120 --> 00:32:00.237
is like okay because these questions
right this question heads towards is

00:32:00.320 --> 00:32:02.637
like okay the impact of social media
and the undermining of truth and all this

00:32:02.720 --> 00:32:05.277
but then you want to ask the question
of like okay what if we had had the modern

00:32:05.360 --> 00:32:08.637
media environment including cable news
and including social media and Twitter

00:32:08.720 --> 00:32:12.187
and everything else in 1939 or 1941

00:32:12.270 --> 00:32:15.757
[snorts] right or 1910 or 1865 or 1850

00:32:15.840 --> 00:32:20.237
or 1776 >> right um and like I think

00:32:20.320 --> 00:32:24.077
>> you you just introduced like five
thought experiments at once and broke my

00:32:24.160 --> 00:32:27.357
head But yes, >> that's there's
a lot of interesting years in

00:32:27.440 --> 00:32:30.077
>> Kennedy like can I just take a simple
example Kennedy like how would President

00:32:30.160 --> 00:32:34.157
Kennedy have been interpreted with what
we know now about all the things Kennedy

00:32:34.240 --> 00:32:36.877
was up to >> like how would
he have been experienced

00:32:36.960 --> 00:32:38.957
by the body politic in a in with a

00:32:39.040 --> 00:32:41.677
social media context right like how

00:32:41.760 --> 00:32:44.957
would LBJ have been experienced um by

00:32:45.040 --> 00:32:48.077
the way how would you know like many FDR

00:32:48.160 --> 00:32:51.997
like the New Deal the great depression >>
I wonder where Twitter would would think

00:32:52.080 --> 00:32:56.717
about Churchill and Hitler and Stalin,
>> you know, I mean, look, to this day

00:32:56.800 --> 00:32:59.517
there, you know, there's there are
lots of very interesting real questions

00:32:59.600 --> 00:33:02.237
around like how America, you know, got,
you know, basically involved in World

00:33:02.320 --> 00:33:05.357
War II and who did what when and the
operations of British intelligence and

00:33:05.440 --> 00:33:08.957
American soil and did FDR
this that Pearl Harbor, you know.

00:33:09.040 --> 00:33:11.277
>> Yeah. >> Where Wilson
ran for, you know, his his

00:33:11.360 --> 00:33:13.997
his candidacy was run on an anti-war
will, you know, this he ran on the

00:33:14.080 --> 00:33:16.557
platform and not getting involved
in World War I. Somehow that switched, you

00:33:16.640 --> 00:33:19.357
know, like, and I'm not even making
a value judgement any of these things. I'm

00:33:19.440 --> 00:33:23.917
just saying like we we the way that
our ancestors experienced reality was of

00:33:24.000 --> 00:33:28.397
course mediated through centralized top
down right control. At that point if you

00:33:28.480 --> 00:33:32.797
if you ran those realities again with
the media environment we have today the

00:33:32.880 --> 00:33:35.517
reality would the reality would be
experienced very very differently and

00:33:35.600 --> 00:33:39.197
then of course that that intermediation
would cause the feedback loops to change

00:33:39.280 --> 00:33:40.877
and then reality would obviously play out.

00:33:40.960 --> 00:33:42.637
>> You think you think it
would be very different?

00:33:42.720 --> 00:33:45.837
>> Yeah, it has to be. It has to be just
because it's all so I mean just look at

00:33:45.920 --> 00:33:48.717
what's happening today. I mean just I
mean the most obvious thing is just the

00:33:48.800 --> 00:33:52.077
the collapse and here's [clears throat]
another opportunity to argue that this

00:33:52.160 --> 00:33:55.197
is not the internet causing this by
the way. Um here's a big thing happening

00:33:55.280 --> 00:33:58.077
today which is Gallup does this thing
every year where they do they pull for

00:33:58.160 --> 00:34:01.037
trust in institutions in [clears throat]
America and they do it across all the

00:34:01.120 --> 00:34:04.077
everything from the military to clergy
and big business and the the media and

00:34:04.160 --> 00:34:07.037
so forth, right? >> Um
and basically there's been a systemic

00:34:07.120 --> 00:34:11.197
collapse um in trust in institutions
in the US almost without exception

00:34:11.280 --> 00:34:16.077
basically since essentially the early
1970s. Um there's two ways of looking at

00:34:16.160 --> 00:34:18.877
that which is oh my god we've lost
this old world in which we could trust

00:34:18.960 --> 00:34:21.437
institutions and that was so much better
because like that should be the way the

00:34:21.520 --> 00:34:24.557
world runs. The other way of looking
at it is we just know a lot more now and

00:34:24.640 --> 00:34:27.277
the great mystery is why
those numbers aren't all zero.

00:34:27.360 --> 00:34:29.197
>> Yeah. [clears throat] >> Right.
[laughter] Because like now we

00:34:29.280 --> 00:34:31.197
know so much about how these things
operate and like they're not that

00:34:31.280 --> 00:34:34.797
impressive. >> And also why
do we don't have uh better

00:34:34.880 --> 00:34:36.557
institutions and better leaders then?

00:34:36.640 --> 00:34:39.517
>> Yeah. And so so so this goes to the
thing which is like okay had had we had

00:34:39.600 --> 00:34:43.357
the media environment of the of that we've
had between the 1970s and today if

00:34:43.440 --> 00:34:47.917
we had that in the 30s and 40s or 1900s
1910s I think there's no question

00:34:48.000 --> 00:34:50.797
reality would turn out different if only
because everybody would have known to

00:34:50.880 --> 00:34:54.397
not trust the institutions which would
have changed their level of credibility

00:34:54.480 --> 00:34:57.437
their ability to control circumstances
therefore the circumstances would have

00:34:57.520 --> 00:35:00.397
had to change right and it would have
been a feedback loop it was it would

00:35:00.480 --> 00:35:03.437
have been a feedback loop process
in other words right it's it's it's it's

00:35:03.520 --> 00:35:07.037
your exper your experience of reality
changes reality and then reality changes

00:35:07.120 --> 00:35:09.997
your experience of reality, right? It's
it's a it's a two-way feedback process

00:35:10.080 --> 00:35:13.997
and media is the intermediating force
>> between that. So change the media

00:35:14.080 --> 00:35:15.757
environment, change reality. >> Yeah.

00:35:15.840 --> 00:35:19.277
>> And so it's just so just as a as a
consequence, I think it's just really

00:35:19.360 --> 00:35:22.797
hard to say, oh, things worked a certain
way then and they work a different way

00:35:22.880 --> 00:35:25.277
now and then therefore like people were

00:35:25.360 --> 00:35:30.077
smarter then or better then or,
you know, by the way dumber then or not as

00:35:30.160 --> 00:35:34.557
capable then, right? We we make all
these like really light and casual like

00:35:34.640 --> 00:35:37.917
comparisons of ourselves to, you know,
previous generations of people. You we

00:35:38.000 --> 00:35:40.557
draw judgments all the time. And I just
think it's like really hard to do any of

00:35:40.640 --> 00:35:43.837
that because if we if we put ourselves
in their shoes with the media that they

00:35:43.920 --> 00:35:46.877
had at that time, like I think we
probably most likely would have been

00:35:46.960 --> 00:35:47.917
just like them.

00:35:48.000 --> 00:35:51.117
>> So don't you think that our perception

00:35:51.200 --> 00:35:54.157
and understanding of reality would you

00:35:54.240 --> 00:35:58.797
be more and more mediated through large
language models now? So you said media

00:35:58.880 --> 00:36:02.957
before. Isn't the LLM
going to be the new what

00:36:03.040 --> 00:36:05.277
is it? Mainstream media MSM. It'll be

00:36:05.360 --> 00:36:09.997
LLM. Uh that would be the source of uh I'm

00:36:10.080 --> 00:36:13.997
sure there's a way to kind of rapidly
fine-tune like making LLM's real time.

00:36:14.080 --> 00:36:18.477
I'm sure there's probably a research
problem that you can uh do just rapid

00:36:18.560 --> 00:36:21.517
fine-tuning to the new events.
Something like this.

00:36:21.600 --> 00:36:25.117
>> Well, even just the the whole concept
of the chat UI might not be the like the

00:36:25.200 --> 00:36:27.677
chat UI is just the first whack at this
and maybe that's the dominant thing. But

00:36:27.760 --> 00:36:30.397
look, may maybe maybe our maybe we
don't we don't know yet. Like maybe the

00:36:30.480 --> 00:36:34.157
experience most people have with LM
is is just a continuous feed, you know,

00:36:34.240 --> 00:36:36.797
maybe it's more of a passive feed
and you just are getting a constant like

00:36:36.880 --> 00:36:39.037
running commentary on everything
happening in your life and it's just

00:36:39.120 --> 00:36:42.157
helping you kind of interpret and
understand everything. also really more

00:36:42.240 --> 00:36:46.717
deeply integrated into your life.
Not just like oh uh like intellectual

00:36:46.800 --> 00:36:49.597
philosophical thoughts but like literally

00:36:49.680 --> 00:36:52.317
uh like how to make a coffee, where to

00:36:52.400 --> 00:36:57.277
go for lunch, just uh whether you
know how dating all this kind of stuff.

00:36:57.360 --> 00:36:58.397
>> What to say in a job interview? Yeah.

00:36:58.480 --> 00:37:01.277
>> What to say?
What to say >> next sentence?

00:37:01.360 --> 00:37:04.317
>> Yeah. Next sentence. Yeah. At that
level. Yeah. I mean, yes. So technically

00:37:04.400 --> 00:37:07.597
now whether we want that or not is an
open question, right? Boy, I would kill

00:37:07.680 --> 00:37:10.077
for a popup. [laughter] A popup. Right

00:37:10.160 --> 00:37:13.197
now, the estimated engagement using is

00:37:13.280 --> 00:37:15.437
decreasing. For Mark Andre since there's

00:37:15.520 --> 00:37:17.677
this controversy uh section for his

00:37:17.760 --> 00:37:22.557
Wikipedia page in 1993, something
happened or something like this. Bring

00:37:22.640 --> 00:37:25.117
it up. That'll drive engagement up anyway.

00:37:25.200 --> 00:37:27.197
>> Yeah, that's right.
I mean, look, [laughter]

00:37:27.280 --> 00:37:29.677
this gets this whole thing of like,
so you know, the chat interface has this

00:37:29.760 --> 00:37:32.797
whole concept of prompt engineering,
right? So prompts. Well, it turns out

00:37:32.880 --> 00:37:36.877
one of the things that OM are
really good at is writing prompts.

00:37:36.960 --> 00:37:40.877
Right. And so like what if you just
outsourced and and by the way you could

00:37:40.960 --> 00:37:42.877
run this experiment today. You could
hook this up to do this today. The

00:37:42.960 --> 00:37:45.277
latency is not good enough to do it real
time in a conversation, but you could

00:37:45.360 --> 00:37:48.557
you could run this experiment and you
just say look every 20 seconds you could

00:37:48.640 --> 00:37:52.557
just say you know you know tell me
what the optimal prompt is and then ask

00:37:52.640 --> 00:37:54.557
yourself that question
to give me the result.

00:37:54.640 --> 00:37:58.237
>> Um and then as as you exactly to
your point as you add there will be there

00:37:58.320 --> 00:38:00.797
will be these systems are going to have
the ability to be learned and updated

00:38:00.880 --> 00:38:03.917
essentially in real time. And so you'll
be able to have a pendant or your phone

00:38:04.000 --> 00:38:06.797
or whatever watch or whatever. It'll
have a microphone on it. It'll listen to

00:38:06.880 --> 00:38:09.837
your conversations. It'll have a feed
of everything else happen in the world and

00:38:09.920 --> 00:38:12.477
then it'll be re, you know, sort
of retraining, prompting or retraining

00:38:12.560 --> 00:38:15.677
itself on the fly. Um, and so the
scenario you described is a is actually

00:38:15.760 --> 00:38:19.677
a completely doable scenario. Now, the
hard question on these is always, okay,

00:38:19.760 --> 00:38:22.637
since that's possible, are people going
to want that? Like, what's the form of

00:38:22.720 --> 00:38:25.357
experience? >> Mh. >> You know,
that that we we won't know

00:38:25.440 --> 00:38:30.077
until we try it. But I don't think it's
possible yet to predict the form of AI

00:38:30.160 --> 00:38:32.957
in our lives. Therefore, it's not
possible to predict the way in which it

00:38:33.040 --> 00:38:35.997
will intermediate
our experience with reality yet.

00:38:36.080 --> 00:38:40.237
>> Yeah. But it feels like there's going
to be a killer app. There's probably a mad

00:38:40.320 --> 00:38:42.157
scramble right now. It's out open AI and

00:38:42.240 --> 00:38:45.117
Microsoft and Google and Meta and then

00:38:45.200 --> 00:38:47.677
startups and smaller companies figuring

00:38:47.760 --> 00:38:52.797
out what is the killer app because it
feels like it's possible like a chat GPT

00:38:52.880 --> 00:38:54.637
type of thing. It's possible to build

00:38:54.720 --> 00:38:57.517
that, but that's 10x more compelling

00:38:57.600 --> 00:39:00.397
using already the LLMs we have using

00:39:00.480 --> 00:39:02.877
even the open source LLMs, Llama and the

00:39:02.960 --> 00:39:04.557
different variants.

00:39:04.640 --> 00:39:09.917
Um, so you're investing in a lot
of companies and you're paying attention.

00:39:10.000 --> 00:39:11.277
Who do you think is going to win this?

00:39:11.360 --> 00:39:16.557
You think they'll be who's going
to be the next page rank inventor?

00:39:16.640 --> 00:39:19.757
>> Trillion dollar question. Um, >>
another one. We have a few of those today.

00:39:19.840 --> 00:39:22.797
>> A bunch of those. So, look, there's
a really big question today. sitting here

00:39:22.880 --> 00:39:26.157
today is a really big question about
the big models versus the small models. Um

00:39:26.240 --> 00:39:30.157
that's related directly to the big
question of proprietary versus open.

00:39:30.240 --> 00:39:33.757
>> Um then there's this big question of
of of you know where is the training data

00:39:33.840 --> 00:39:36.557
going to like are we topping out on
the training data or not and then are we

00:39:36.640 --> 00:39:40.077
going to be able to synthesize training
data and then there's a huge pile of

00:39:40.160 --> 00:39:43.597
questions around regulation um and you
know what's actually going to be legal

00:39:43.680 --> 00:39:47.357
um and so I would I when we think about
it we we dovetail kind of all those all

00:39:47.440 --> 00:39:51.117
those questions together. You can paint
a picture of the world where there's two

00:39:51.200 --> 00:39:53.437
or three god models that are just at

00:39:53.520 --> 00:39:58.237
like staggering scale. Um, and they're
just better at everything. Um, and they

00:39:58.320 --> 00:40:00.877
will be owned by a small set
of companies and they will basically

00:40:00.960 --> 00:40:03.437
achieve regulatory capture over the
government and they'll have competitive

00:40:03.520 --> 00:40:06.237
barriers that will prevent other people
from um, you know, competing with them.

00:40:06.320 --> 00:40:08.797
And so, you know, there will be, you know,
just like there's like, you know,

00:40:08.880 --> 00:40:11.837
whatever three big banks or three big
you or by the way, three big search

00:40:11.920 --> 00:40:16.077
companies or I guess do you know it'll
it'll centralize like that. um you can

00:40:16.160 --> 00:40:19.757
paint another very different picture
that says no um actually the opposite of

00:40:19.840 --> 00:40:23.517
that's going to happen. this is going to
basically that this is the new gold, you

00:40:23.600 --> 00:40:27.677
know, this is the new gold rush alchemy
like, you know, this is the this is the

00:40:27.760 --> 00:40:31.277
big bang for this whole new area
of of science and technology. And so

00:40:31.360 --> 00:40:33.677
therefore, you're going to have every
smart 14-year-old on the planet building

00:40:33.760 --> 00:40:36.557
open source, right? You know,
and figuring out ways to optimize these

00:40:36.640 --> 00:40:39.917
things. Um, and then, you know, we're
just going to get like overwhelmingly

00:40:40.000 --> 00:40:42.637
better at generating trading data.
We're going to, you know, bring in like

00:40:42.720 --> 00:40:45.037
blockchain networks to have like
an economic incentive to generate

00:40:45.120 --> 00:40:48.797
decentralized training data and so forth
and so on. And then basically, we're

00:40:48.880 --> 00:40:50.077
going to live in a world of open source.

00:40:50.160 --> 00:40:54.157
And there's going to be a billion LLMs,
right, of every size, scale, shape, and

00:40:54.240 --> 00:40:56.877
description. And there might be a few
big ones that are like the super genius

00:40:56.960 --> 00:40:59.917
ones, but like mostly what we'll
experience is open source. And that's,

00:41:00.000 --> 00:41:02.557
you know, that's more like a world
of like what we have today with like Linux

00:41:02.640 --> 00:41:05.277
and the web. >> Um, so

00:41:05.360 --> 00:41:09.997
>> okay. But, uh, you you painted these
two worlds, but there's also variations of

00:41:10.080 --> 00:41:12.557
those worlds because you said regulatory
capture. it's possible to have these

00:41:12.640 --> 00:41:16.237
tech giants that don't have regulatory
capture, which is something you're also

00:41:16.320 --> 00:41:20.237
calling for, saying it's okay to have
big companies working on this stuff, uh,

00:41:20.320 --> 00:41:21.917
as long as they don't achieve regulatory

00:41:22.000 --> 00:41:27.117
capture. Uh, but I have the sense that,

00:41:27.200 --> 00:41:30.557
um, there's just going to be a new startup

00:41:30.640 --> 00:41:33.197
that's going to basically be the page

00:41:33.280 --> 00:41:38.397
rank inventor, which
is become the new tech giant.

00:41:38.480 --> 00:41:40.077
I don't know. I would love to hear your

00:41:40.160 --> 00:41:43.837
kind of opinion if Google, Meta, and

00:41:43.920 --> 00:41:47.677
Microsoft are as gigantic companies able

00:41:47.760 --> 00:41:51.277
to pivot so hard to create new products.

00:41:51.360 --> 00:41:55.037
Like some of it is just even hiring
people or having uh corporate structure

00:41:55.120 --> 00:41:59.437
that allows for the crazy young kids
to come in and just create something

00:41:59.520 --> 00:42:02.957
totally new. Do you think it's possible
or do you think it'll come from a startup?

00:42:03.040 --> 00:42:05.917
>> Yeah, it is this always big question
which is you get this feeling. I hear

00:42:06.000 --> 00:42:10.157
about this a lot from CEOs, founder
CEOs where it's like, wow, we have 50,000

00:42:10.240 --> 00:42:13.757
people. It's now harder to do new
things than it was when we had 50 people.

00:42:13.840 --> 00:42:15.517
>> Yeah. >> Like what has
happened? So that that's a

00:42:15.600 --> 00:42:18.637
recurring phenomenon. Um, by the way,
that's one of the reasons why there's

00:42:18.720 --> 00:42:21.597
always startups and why there's venture
capital. Um, it's just that's that's

00:42:21.680 --> 00:42:24.557
like a timeless >> uh kind
of thing. So that that that's

00:42:24.640 --> 00:42:29.597
one observation. Um, on on page rank
um we could talk about that, but on page

00:42:29.680 --> 00:42:32.397
rank specifically on page rank um there
actually is a page so there is a page

00:42:32.480 --> 00:42:35.357
rank already in the field and it's
the transformer, right? So the the the big

00:42:35.440 --> 00:42:39.837
breakthrough was the transformer. Um
and uh the transformer was invented in uh

00:42:39.920 --> 00:42:43.517
2017 >> at Google.
And this is actually like

00:42:43.600 --> 00:42:46.957
really an interesting question because
it's like okay the transformers like why

00:42:47.040 --> 00:42:49.917
does open AAI even exist? Like
the transformers invested at Google, why

00:42:50.000 --> 00:42:53.037
didn't Google I asked a guy I asked
a guy I know who was senior at Google

00:42:53.120 --> 00:42:54.157
Brain kind of when this was happening.

00:42:54.240 --> 00:42:56.957
And I said if Google had
just gone flat out to the wall

00:42:57.040 --> 00:42:58.317
>> and just said look
we're going to launch

00:42:58.400 --> 00:43:01.677
we're going to launch equivalent of GPT4
as fast as we can. Um he said I said

00:43:01.760 --> 00:43:04.317
when could we have had
it and he said 2019. Yeah,

00:43:04.400 --> 00:43:06.957
>> they could have just done a two-year
sprint with the transformer and and been

00:43:07.040 --> 00:43:09.277
at because they already had the compute
at scale. They already had all the

00:43:09.360 --> 00:43:11.037
training data.
They could have just done it.

00:43:11.120 --> 00:43:14.317
>> There's a variety of reasons they
didn't do it. This is like a classic big

00:43:14.400 --> 00:43:18.797
company thing. Um IBM invented the
relational database in 19 in the 1970s,

00:43:18.880 --> 00:43:20.397
let it sit on the shelf as a paper.

00:43:20.480 --> 00:43:23.837
Larry Ellison picked it up and built
Oracle. Xerox Park invented the

00:43:23.920 --> 00:43:26.877
interactive computer. They let it sit
on the shelf. Steve Jobs came and turned

00:43:26.960 --> 00:43:31.117
into the Macintosh. Right? And so there
is this pattern. Now having said that,

00:43:31.200 --> 00:43:34.317
sitting here today like Google's in the
game, right? So Google, you know, maybe

00:43:34.400 --> 00:43:37.197
maybe they they maybe they let like
a four-year gap there go there that they

00:43:37.280 --> 00:43:39.837
maybe shouldn't have, but like they're
in the game. And so now they've got, you

00:43:39.920 --> 00:43:42.077
know, now they're committed. They've
done this merger. They're bringing in

00:43:42.160 --> 00:43:43.997
Demos. They've got
this merger with Deep Mind.

00:43:44.080 --> 00:43:45.437
>> You know, they're piling in resources.

00:43:45.520 --> 00:43:48.557
There are rumors that they're, you know,
building an incredible, you know, super

00:43:48.640 --> 00:43:53.117
LLM. Um, you know, way beyond what we
even have today. Um, and they've got,

00:43:53.200 --> 00:43:56.157
you know, unlimited resources and a huge,
you know, they've been challenged

00:43:56.240 --> 00:43:59.117
their honor. [laughter] >> Yeah.
I had a I had a a chance to hang

00:43:59.200 --> 00:44:03.037
out with Sund Bachai a couple days ago
and we took this walk and there's this

00:44:03.120 --> 00:44:07.357
giant new building uh where there's
going to be a lot of AI work uh being

00:44:07.440 --> 00:44:09.677
done and it's kind of

00:44:09.760 --> 00:44:15.357
this ominous feeling of like
the fight is on. [laughter]

00:44:15.440 --> 00:44:18.557
>> Yeah. >> Like there's
this beautiful Silicon

00:44:18.640 --> 00:44:23.197
Valley nature like birds are chirping
and this giant building and it's like uh

00:44:23.280 --> 00:44:25.837
the beast has been awakened. >> Yeah.

00:44:25.920 --> 00:44:29.917
And then like all the big companies
are waking up to this. They have the

00:44:30.000 --> 00:44:35.197
compute, >> but also
the little guys have uh it

00:44:35.280 --> 00:44:38.717
feels like they have all the tools
to create the killer product that uh and

00:44:38.800 --> 00:44:42.557
then there's also the tools to scale.
If you have a good idea, if you have the

00:44:42.640 --> 00:44:45.597
page rank idea. So I there's several

00:44:45.680 --> 00:44:49.997
things that is page rank p there's
page rank the algorithm and the idea and

00:44:50.080 --> 00:44:53.357
there's like the implementation of it
and I feel like killer product is not

00:44:53.440 --> 00:44:56.877
just the idea like the transform it's
the implementation something something

00:44:56.960 --> 00:45:01.357
really compelling about it like you just
can't look away something like um the

00:45:01.440 --> 00:45:05.517
algorithm behind Tik Tok versus Tik Tok
itself like the actual experience of Tik

00:45:05.600 --> 00:45:09.597
Tok that just you can't look away.
It feels like somebody's going to come up

00:45:09.680 --> 00:45:12.557
with that >> and it could
be Google, but it feels

00:45:12.640 --> 00:45:16.317
like it's just easier
and faster to do for a startup.

00:45:16.400 --> 00:45:19.437
>> Yeah. So, so the startup the the huge
the huge advantage that startups have is

00:45:19.520 --> 00:45:21.197
they just they there's no sacred cows.

00:45:21.280 --> 00:45:22.957
There's no historical legacy to protect.

00:45:23.040 --> 00:45:25.757
There's no need to reconcile your new
plan with the existing strategy. There's

00:45:25.840 --> 00:45:28.797
no communication overhead. There's no,
you know, big companies are big

00:45:28.880 --> 00:45:31.517
companies. They've got pre meetings
planning for the meeting. Then they have

00:45:31.600 --> 00:45:33.677
then they have the post meeting of the
recap. Then they have the presentation

00:45:33.760 --> 00:45:35.437
of the board. Then they have
the next rounds of meetings.

00:45:35.520 --> 00:45:38.477
>> Yeah. Yeah. And and that's that's
the elapsed time when the startup launches

00:45:38.560 --> 00:45:42.797
its product, right? So so so so there's
a timeless, right? So there's a timeless

00:45:42.880 --> 00:45:45.037
thing there. Now >> what
the startups don't have is

00:45:45.120 --> 00:45:47.357
everything else, right? So startups,
they don't have a brand, they don't have

00:45:47.440 --> 00:45:49.917
customer relationships, they've got no
distribution, they've got no, you know,

00:45:50.000 --> 00:45:52.717
scale. I mean, sitting here today, they
can't even get GPUs, right? Like there's

00:45:52.800 --> 00:45:55.597
like a GPU shortage. >> Startups
are literally stalled out right

00:45:55.680 --> 00:45:58.237
now because they can't get chips,
which is like super weird.

00:45:58.320 --> 00:46:00.317
>> Yeah. >> Um >> they got the cloud.

00:46:00.400 --> 00:46:01.837
>> Yeah. But the clouds run out of chips.

00:46:01.920 --> 00:46:05.117
Um right. And then and then and then
to the extent the clouds have chips. They

00:46:05.200 --> 00:46:08.637
allocate them to the big customers, not
the small customers. Right. And so so so

00:46:08.720 --> 00:46:12.877
so the small companies lack everything
other than the ability to just do

00:46:12.960 --> 00:46:15.917
something new. >> Yeah. >> Right.
Um and and this is the timeless

00:46:16.000 --> 00:46:18.237
race and battle. And this is kind
of the point I tried to make in the essay,

00:46:18.320 --> 00:46:21.517
which is like both sides of this are good.
Like it's really good to have like

00:46:21.600 --> 00:46:24.477
highly scaled tech companies that can
do things that are like at staggering

00:46:24.560 --> 00:46:27.357
levels of sophistication. It's really
good to have startups that can launch

00:46:27.440 --> 00:46:30.637
brand new ideas. They ought to be
able to both do that and compete. they

00:46:30.720 --> 00:46:33.917
neither one ought to be subsidized
or protected from the others. Like that's

00:46:34.000 --> 00:46:37.677
that's to me that's just like very
clearly the idealized world. It is the

00:46:37.760 --> 00:46:39.677
world we've been in for AI up until now.

00:46:39.760 --> 00:46:42.317
And then of course there are people
trying to shut that down. But my hope is

00:46:42.400 --> 00:46:45.037
that you know the best outcome
clearly will be if that continues.

00:46:45.120 --> 00:46:49.277
>> We'll talk about that a little
bit but I'd love to linger uh

00:46:49.360 --> 00:46:53.757
on uh some of the ways this is going
to change the internet. So um I don't know

00:46:53.840 --> 00:46:56.477
if you remember but there's a thing
called Mosaic and there's a thing called

00:46:56.560 --> 00:47:00.557
Netscape Navigator. So you were there
in the beginning. Uh what about the

00:47:00.640 --> 00:47:04.397
interface to the internet? How do you
think the browser changes and who gets

00:47:04.480 --> 00:47:06.077
to own the browser? We got to see some

00:47:06.160 --> 00:47:09.757
very interesting browsers. Uh Firefox, I

00:47:09.840 --> 00:47:12.077
mean all the variants of Microsoft

00:47:12.160 --> 00:47:15.277
Internet Explorer, Edge and uh now

00:47:15.360 --> 00:47:20.557
Chrome. Um the actual
I mean it seems like a

00:47:20.640 --> 00:47:24.717
dumb question to ask, but do you
think we'll still have the web browser?

00:47:24.800 --> 00:47:28.477
So, I uh I have an 8-year-old and he's
super into like Minecraft and learning

00:47:28.560 --> 00:47:31.677
to code and doing all this stuff. So,
I I I of course I was very proud I could

00:47:31.760 --> 00:47:34.157
bring sort of fire down from the
mountain to my kid and I brought him

00:47:34.240 --> 00:47:36.157
chat GPT and I hooked him up >> Yeah.

00:47:36.240 --> 00:47:39.197
>> on his on his on his on his laptop and
I was like, you know, this is the thing

00:47:39.280 --> 00:47:41.757
that's going to answer all your questions.
And he's like, okay. [snorts]

00:47:41.840 --> 00:47:44.397
>> And I'm like, but it's going to answer
all your questions. And he's like, well,

00:47:44.480 --> 00:47:46.397
of course, like it's a computer.
Of course it answers all your questions.

00:47:46.480 --> 00:47:49.437
Like, what else would
a computer be good for? Dad.

00:47:49.520 --> 00:47:51.437
>> Um, [snorts] and never impressed.

00:47:51.520 --> 00:47:55.837
>> Not impressed in the least. Two weeks
pass. Um, and he has some question. Um,

00:47:55.920 --> 00:47:59.117
and I say, "Well, have you asked Jet GPT?"
And he's like, "Dad, Bing is

00:47:59.200 --> 00:48:02.877
better." >> And why is Bing
better is because it's

00:48:02.960 --> 00:48:06.157
built into the browser? Because he's
like, "Look, I have the Microsoft Edge

00:48:06.240 --> 00:48:08.717
browser and like it's got Bing right
here." And then he he doesn't know this

00:48:08.800 --> 00:48:11.997
yet, but one of the things you can
do with Bing in Edge um is there's a

00:48:12.080 --> 00:48:16.397
setting where you can um use it
to basically talk to any web page because

00:48:16.480 --> 00:48:19.197
it's sitting right there next to the uh
next to the next to the browser. And by

00:48:19.280 --> 00:48:22.317
the way, which includes PDF documents
and so you can in in in the way they've

00:48:22.400 --> 00:48:25.437
implemented in Edge with Bang is you can
load a PDF and then you can you can ask

00:48:25.520 --> 00:48:28.877
it questions which is the thing you you
you can't do currently and and and just

00:48:28.960 --> 00:48:31.837
chat GPT. So they're you know they're
they're going to they're going to push

00:48:31.920 --> 00:48:34.317
the the the mel I think that's great you
know they're going to push the melding

00:48:34.400 --> 00:48:37.597
and see if there's a combination thing
there. Google's rolling out this thing

00:48:37.680 --> 00:48:41.517
the magic button which is implemented
in they put in Google Docs, right? And so

00:48:41.600 --> 00:48:44.797
you go into, you know, Google Docs
and you create a new document and you, you

00:48:44.880 --> 00:48:47.117
know, you instead of like, you know,
starting to type, you just, you know,

00:48:47.200 --> 00:48:50.877
say it, press the button and it starts
to like generate content for you, right?

00:48:50.960 --> 00:48:53.037
Like is that the way that it'll work?

00:48:53.120 --> 00:48:56.557
Um, is it going to be a speech UI where
you're just going to have an earpiece

00:48:56.640 --> 00:49:00.797
and talk to it all day long? You know,
is it going to be a like these are all

00:49:00.880 --> 00:49:03.597
like this is exactly the kind of thing
that I I don't this is exactly the kind

00:49:03.680 --> 00:49:06.397
of thing I don't think is possible
to forecast. I think what we need to do is

00:49:06.480 --> 00:49:09.837
like run all those experiments. Um and
and and so one outcome is we come out of

00:49:09.920 --> 00:49:13.347
this with like a super browser that has
AI built in that's just like amazing.

00:49:13.430 --> 00:49:15.997
[snorts] the other there. Look, there's
a real possibility that the whole I

00:49:16.080 --> 00:49:19.677
mean, look, there's a possibility here
that the whole idea of a screen and

00:49:19.760 --> 00:49:22.957
windows and all this stuff just goes
away because like why do you need that

00:49:23.040 --> 00:49:26.317
if you just have a thing that's just
telling you whatever you need to know?

00:49:26.400 --> 00:49:29.277
>> And also, so there's apps that you can

00:49:29.360 --> 00:49:31.277
use. You don't really use them, you

00:49:31.360 --> 00:49:34.317
know, being a Linux guy and Windows guy.

00:49:34.400 --> 00:49:38.797
Um, there's one window, the browser
that with which you can interact with the

00:49:38.880 --> 00:49:42.397
internet, but on the phone, you can
also have apps. So, I can interact with

00:49:42.480 --> 00:49:47.197
Twitter through the app or through
the web browser and um that seems like an

00:49:47.280 --> 00:49:51.197
obvious distinction, but why have
the web browser in that case if one of the

00:49:51.280 --> 00:49:53.837
apps starts becoming the everything app?

00:49:53.920 --> 00:49:55.757
>> Yeah, that's right.
>> What Elon's trying to do with Twitter,

00:49:55.840 --> 00:49:58.877
but there could be others. There could
be like a Bing app. There could be a

00:49:58.960 --> 00:50:03.597
Google app that just doesn't
really do search, but just like

00:50:03.680 --> 00:50:05.437
do what I guess AOL did back in the day

00:50:05.520 --> 00:50:08.957
or something where it's all right there

00:50:09.040 --> 00:50:11.997
and and it changes

00:50:12.080 --> 00:50:15.037
um it changes the nature of the internet

00:50:15.120 --> 00:50:17.197
because

00:50:17.280 --> 00:50:22.157
the where the content is hosted, who
owns the data, who owns the content, how

00:50:22.240 --> 00:50:25.517
what is what is the kind of content
you create? How do you make money by

00:50:25.600 --> 00:50:28.877
creating content? or the content creators,

00:50:28.960 --> 00:50:33.837
uh, all of that. Or it could just keep
being the same, which is like with just

00:50:33.920 --> 00:50:36.957
the nature of web pages changes and the
nature of content, but there will still

00:50:37.040 --> 00:50:40.077
be a web browser cuz a web
browser is a pretty sexy product.

00:50:40.160 --> 00:50:41.597
>> Mhm. >> It just seems to work.

00:50:41.680 --> 00:50:44.317
>> Mhm. >> Cuz it like
you have an interface, a

00:50:44.400 --> 00:50:47.597
window into the world, and then
the world can be anything you want. And as

00:50:47.680 --> 00:50:50.397
the world will evolve, it could be
different programming languages, it can

00:50:50.480 --> 00:50:54.077
be animated, maybe it's threedimensional,
and so on. Yeah, it's

00:50:54.160 --> 00:50:56.957
interesting. Do you think we'll
still have the web browser?

00:50:57.040 --> 00:51:01.277
>> Every every every um every medium
becomes the content for the next one. So

00:51:01.360 --> 00:51:04.477
the you know the AI will be able to give
you a browser whenever you want. Um

00:51:04.560 --> 00:51:07.117
>> Oh, interesting. >> Yeah.
Well, another way to think about

00:51:07.200 --> 00:51:10.717
it is maybe what the browser is, maybe
it's just the escape hatch, right? And

00:51:10.800 --> 00:51:14.317
which is maybe kind of what it is today,
right? Which is like most of what you do

00:51:14.400 --> 00:51:17.197
is like inside a social network
or inside a search engine or inside, you

00:51:17.280 --> 00:51:20.397
know, somebody's app or inside some
controlled experience, right? But then

00:51:20.480 --> 00:51:23.357
every once in a while there's something
where you actually want to jailbreak.

00:51:23.440 --> 00:51:24.557
you want to actually get free.

00:51:24.640 --> 00:51:26.877
>> The web browser
is the f you to the man.

00:51:26.960 --> 00:51:29.357
You you're allowed
to That's the free internet.

00:51:29.440 --> 00:51:31.837
>> Yeah. >> Back back
the way it was in the '90s.

00:51:31.920 --> 00:51:33.677
>> So, here's something I'm proud of.
So, nobody really talks about here's

00:51:33.760 --> 00:51:35.917
something I'm proud of, which is that
the web the web the browser the web

00:51:36.000 --> 00:51:37.997
servers they're all they're still
backward compatible all the way back to

00:51:38.080 --> 00:51:43.037
like 1992, right? So, like you can
put up a you can still you know the big

00:51:43.120 --> 00:51:45.837
breakthrough of the web early on the big
breakthrough was it made it really easy

00:51:45.920 --> 00:51:48.717
to read, but it also made it really easy
to write, made it really easy to publish

00:51:48.800 --> 00:51:51.917
and and and we literally made it so easy
to publish. We made it not only so it

00:51:52.000 --> 00:51:54.637
was easy to publish content, it
was actually also easy to actually write a

00:51:54.720 --> 00:51:56.877
web server. >> Yeah. >> Right.
And and you could literally write

00:51:56.960 --> 00:51:59.837
a web server in four lines of Pearl code
and and and you could start publishing

00:51:59.920 --> 00:52:02.957
content on it and you could set whatever
rules you want for the content, whatever

00:52:03.040 --> 00:52:06.077
censorship, no censorship, whatever you
want, you could just do that as long as

00:52:06.160 --> 00:52:10.557
you had an IP address, right? You you
could do that. That still works, right?

00:52:10.640 --> 00:52:14.477
Like that still works exactly as I
just described. So this is part of my

00:52:14.560 --> 00:52:17.597
reaction to all of this like you know
all this just censorship pressure and

00:52:17.680 --> 00:52:20.717
all this you know these issues around
control and all this stuff which is like

00:52:20.800 --> 00:52:24.317
maybe we need to get back a little bit
more to the wild west like the wild west

00:52:24.400 --> 00:52:27.757
is still out there now they [laughter]
will they will try to chase you down

00:52:27.840 --> 00:52:30.157
like they'll try to you know people
who want to censor will try to take away

00:52:30.240 --> 00:52:32.797
your your um you know your domain
name and they'll try to take away your

00:52:32.880 --> 00:52:35.437
payments account and so forth if they
really don't like what you what you're

00:52:35.520 --> 00:52:38.797
saying but but nevertheless you like
unless they literally are intercepting

00:52:38.880 --> 00:52:43.117
you at the ISP level like you can still
put up a thing um and so I don't know. I

00:52:43.200 --> 00:52:46.717
think that's important to preserve,
right? Like because because because I

00:52:46.800 --> 00:52:50.077
mean one is just a a freedom argument,
but the other is a creativity argument

00:52:50.160 --> 00:52:52.957
which is you want to have the escape
hatch so that the kid with the idea is

00:52:53.040 --> 00:52:55.837
able to realize the idea because to
your point on page rank you you actually

00:52:55.920 --> 00:52:59.277
don't know what the next big idea is,
right? No, nobody called Larry Page and

00:52:59.360 --> 00:53:01.917
told him to develop page rank like he
came up with that on his own. And you

00:53:02.000 --> 00:53:04.957
want to always I I think leave the escape
hatch for the next, you know, kid

00:53:05.040 --> 00:53:07.597
or the next Stanford grad student
to have the breakthrough idea and be able

00:53:07.680 --> 00:53:11.757
to get it up and running before anybody
notices. Um, you and I are both fans of

00:53:11.840 --> 00:53:13.757
history. So, let's step back. We've been

00:53:13.840 --> 00:53:14.957
talking about the future. Let's step

00:53:15.040 --> 00:53:18.797
back for a bit and look at uh the '9s.

00:53:18.880 --> 00:53:23.357
You created Mosaic web browser, the first
widely used web browser. Tell the

00:53:23.440 --> 00:53:27.677
story of that. How and how did it evolve
into Netscape Navigator. This the early

00:53:27.760 --> 00:53:30.877
days. >> So, full story. So, um

00:53:30.960 --> 00:53:34.717
>> you were born >> I
was born small child. Um

00:53:34.800 --> 00:53:38.317
>> actually, yeah, let's go there.
like when when did you when did you first

00:53:38.400 --> 00:53:42.077
fall in love with computers? >> Oh,
so I hit the generational jackpot

00:53:42.160 --> 00:53:45.037
and I hit the Gen X kind of point
perfectly as it turns out. So I was born

00:53:45.120 --> 00:53:50.477
in 1971. So there's this great website
called WTF happened in 1971.com which is

00:53:50.560 --> 00:53:53.037
basically 1971 is when everything
started to go to hell and I was of

00:53:53.120 --> 00:53:56.237
course born in 1971. So I like to think
that I had something to do with that.

00:53:56.320 --> 00:53:59.757
>> Did you make it on the website? I have
I don't think I made it on the website but

00:53:59.840 --> 00:54:03.277
I you know hopefully >> somebody needs
to add this is this is where everything

00:54:03.360 --> 00:54:07.037
>> maybe I contributed to some of the
trends um that they uh that they should

00:54:07.120 --> 00:54:10.157
every line on that website goes like
that right so it's all it's all it's all

00:54:10.240 --> 00:54:14.157
a picture disaster but um but there
was this moment in time where because the

00:54:14.240 --> 00:54:17.677
you know sort of the Apple you know
the Apple 2 hit in like 1978 and then the

00:54:17.760 --> 00:54:22.637
IBM PC hit in ' 82 so I was like you
know 11 when the PC came out um and so I

00:54:22.720 --> 00:54:25.677
just kind of hit that perfectly and then
that was the first moment in time when

00:54:25.760 --> 00:54:28.477
like regular people could spend a few
hundred and get a computer, right? And

00:54:28.560 --> 00:54:32.717
so that I just like that that that
resonated right out of the gate. Um, and

00:54:32.800 --> 00:54:35.597
then the other part of the story is,
you know, I was using an Apple 2. I used a

00:54:35.680 --> 00:54:37.837
bunch of them, but I was using Apple
2 and of course it said on the back of

00:54:37.920 --> 00:54:41.837
every Apple 2 and every Mac it said,
you know, designed in Certino, California.

00:54:41.920 --> 00:54:45.357
And I was like, wow, Certino must be the
like shiny city on the hill, like Wizard

00:54:45.440 --> 00:54:48.637
of Oz, like the most amazing like city
of all time. I can't wait to see it. And

00:54:48.720 --> 00:54:52.317
of course, years later, I came out
to Silicon Valley and went to Certino and

00:54:52.400 --> 00:54:55.117
it's just a bunch of office parks
[laughter] and lowrise apartment

00:54:55.200 --> 00:54:58.477
buildings. [gasps] So the aesthetics
were a little disappointing but you know

00:54:58.560 --> 00:55:02.717
it was the the vector uh right of the
of the creation of a lot of a lot of this

00:55:02.800 --> 00:55:07.437
stuff. Um >> so so then
basically by so part part of

00:55:07.520 --> 00:55:09.757
my story is just the luck of having
been born at the right time and getting

00:55:09.840 --> 00:55:13.517
exposed to PCs. Then the other part
is um the other part is when El Gore says

00:55:13.600 --> 00:55:16.957
that he created the internet. he
actually is correct uh in in in a really

00:55:17.040 --> 00:55:20.397
meaningful way which is he sponsored
a bill in 1985 that essentially created

00:55:20.480 --> 00:55:24.157
the modern internet created what
is called the NSFnet at the time which is

00:55:24.240 --> 00:55:28.717
sort of the the first really fast
internet backbone. Um and uh you know

00:55:28.800 --> 00:55:31.997
that that build dumped a ton of money
into a bunch of research universities to

00:55:32.080 --> 00:55:34.797
build out basically the internet
backbone and then these supercomputer

00:55:34.880 --> 00:55:38.317
centers that were clustered around
um the the the internet and and one of

00:55:38.400 --> 00:55:41.197
those universities was University
of Illinois where I went to school. And so

00:55:41.280 --> 00:55:43.917
the other stroke of luck that I had was I
I went to Illinois basically right as

00:55:44.000 --> 00:55:47.597
that money was just like getting dumped
on campus. And so as a consequence we

00:55:47.680 --> 00:55:52.797
had at on campus and this is like you know
8991 we had like you know we were right on

00:55:52.880 --> 00:55:56.237
the internet backbone. We had like
T3 and 45 at the time T3 45 megabit

00:55:56.320 --> 00:55:59.597
backbone connection which at the time
was you know wildly state-of-the-art. Um

00:55:59.680 --> 00:56:02.557
we had Cray supercomputers we
had thinking machines parallel

00:56:02.640 --> 00:56:06.237
supercomputers. We had silicon graphics
workstations. We had Macintoshes. We had

00:56:06.320 --> 00:56:09.517
we had next cubes all over the place. We
had like every possible kind of computer

00:56:09.600 --> 00:56:13.037
you could imagine cuz all this money
just fell out of the sky. [gasps] Um,

00:56:13.120 --> 00:56:14.557
>> so you were living in the future.

00:56:14.640 --> 00:56:15.997
>> Yeah. So, yeah, quite literally it was.

00:56:16.080 --> 00:56:18.957
Yeah. Like it's all it's all there. It's
all like we had full broadband graphics

00:56:19.040 --> 00:56:21.917
like the whole thing. And and it's
actually funny because they had this

00:56:22.000 --> 00:56:24.957
this is the first time I kind of it
sort of tickled the back of my head that

00:56:25.040 --> 00:56:27.757
there might be a big opportunity in here,
which is, you know, they they

00:56:27.840 --> 00:56:30.157
embraced it and so they put like
computers in all the dorms and they

00:56:30.240 --> 00:56:33.197
wired up all the dorm rooms and they had
all these, you know, labs everywhere and

00:56:33.280 --> 00:56:34.477
everything. And then they they gave

00:56:34.560 --> 00:56:36.957
every undergrad a computer account and

00:56:37.040 --> 00:56:41.677
an email address. Um and the assumption
was that you would use the internet for

00:56:41.760 --> 00:56:46.877
your four years at college. Um and then
you would graduate and stop using it

00:56:46.960 --> 00:56:48.477
>> and that was that right? Yeah.

00:56:48.560 --> 00:56:50.637
>> And you would just retire your email
address. It wouldn't be relevant anymore

00:56:50.720 --> 00:56:53.277
because you'd go off in the workplace
and they don't use email. [snorts]

00:56:53.360 --> 00:56:55.437
>> You'd be back to using
fax machines or whatever.

00:56:55.520 --> 00:56:58.557
>> Did you have that sense as well? Like
what what you said the the back of your

00:56:58.640 --> 00:57:02.157
head was tickled like what what was your
what what was exciting to you about this

00:57:02.240 --> 00:57:05.357
possible world? Well, if this if this is
so useful in this contain if this is so

00:57:05.440 --> 00:57:07.757
useful in this contained environment
that just has this weird source of

00:57:07.840 --> 00:57:11.357
outside funding, then if if it were
practical for everybody else to have

00:57:11.440 --> 00:57:13.837
this and if it were cost- effective
for everybody else to have this, wouldn't

00:57:13.920 --> 00:57:17.517
they want it? And the overwhelmingly the
prevailing view at the time was no, they

00:57:17.600 --> 00:57:20.957
would not want it. This is esoteric
weird nerd stuff, right, that like

00:57:21.040 --> 00:57:23.197
computer science kids like, but like
normal people are never going to do

00:57:23.280 --> 00:57:27.277
email, right, or be on the internet,
right? Um, and so I was just like, wow,

00:57:27.360 --> 00:57:30.497
like this this is actually like
this is really compelling stuff. Now the

00:57:30.580 --> 00:57:33.437
[snorts] other part was it was all
really hard to use and in practice you

00:57:33.520 --> 00:57:36.397
had to be a basically a CS uh you
basically had you had to be a CS

00:57:36.480 --> 00:57:39.757
undergrad or equivalent to actually get
full use of the internet at that point

00:57:39.840 --> 00:57:42.637
um because it was all pretty esoteric
stuff. So then that was the other part

00:57:42.720 --> 00:57:45.837
of the idea which was okay we need
to actually make this easy to use.

00:57:45.920 --> 00:57:51.037
>> So what's involved in creating mosaic
like in creating a graphical interface

00:57:51.120 --> 00:57:53.597
to the internet? >> Yes.
So it was a combination of things.

00:57:53.680 --> 00:57:57.277
So it was like basically the the web
existed in an early sort of described as

00:57:57.360 --> 00:58:00.557
prototype form and by the way
text only at that point.

00:58:00.640 --> 00:58:01.997
>> What did it look like?
What what was the

00:58:02.080 --> 00:58:06.877
web I mean and the key figures like
what was it what was it like? What

00:58:06.960 --> 00:58:09.137
>> made a picture?
>> It looked like Jad GPT actually.

00:58:09.220 --> 00:58:11.117
[laughter] Um it was all text.

00:58:11.200 --> 00:58:14.477
>> Yeah. >> Um and so you
had a textbased web browser.

00:58:14.560 --> 00:58:16.957
>> Well actually the original browser
Tim Tim Berners Lee the original the

00:58:17.040 --> 00:58:19.437
original browser both the original
browser and the server actually ran on

00:58:19.520 --> 00:58:22.077
next next cubes. Mhm.
>> So these were this was you know the

00:58:22.160 --> 00:58:25.037
computer Steve Jobs made during the
interim period when he during the decade

00:58:25.120 --> 00:58:28.317
long interim period when he was not
at Apple [gasps] you know he got fired in

00:58:28.400 --> 00:58:32.717
85 and then came back in 97. So this was
in that interim period where he had this

00:58:32.800 --> 00:58:36.237
company called Next and they made these
literally these computers called cubes

00:58:36.320 --> 00:58:39.357
>> and there's this famous story they were
beautiful but they were uh 12in x 12 in

00:58:39.440 --> 00:58:42.477
x 12 in cubes computers and there's
a famous story about how they could have

00:58:42.560 --> 00:58:46.557
cost half as much if it had
been 12 x 12 x 13 but [laughter]

00:58:46.640 --> 00:58:50.877
Steve was like no like it has to be. So
they were like $6,000 basically academic

00:58:50.960 --> 00:58:54.877
workstations. They had the first city
round drives. Um which were slow. I mean

00:58:54.960 --> 00:58:58.797
it was the computers were all but
unusable. Um they were so slow but they

00:58:58.880 --> 00:59:01.997
were beautiful. >> Okay. Can we
actually just take a tiny tangent there?

00:59:02.080 --> 00:59:03.677
>> Sure. Of course.

00:59:03.760 --> 00:59:07.197
>> The the 12 x 12 x 12 uh that just so

00:59:07.280 --> 00:59:09.757
beautifully encapsulates Steve Jobs idea

00:59:09.840 --> 00:59:13.277
of design. Can you just comment on um

00:59:13.360 --> 00:59:17.117
what you find interesting about Steve
Jobs? what uh about that view of the

00:59:17.200 --> 00:59:21.917
world that dogmatic pursuit
of perfection in how he saw perfection in

00:59:22.000 --> 00:59:24.317
design. >> Yes.
So I guess they say like look he

00:59:24.400 --> 00:59:28.077
was a deep believer I think in a very
deep the way I interpret it I don't know

00:59:28.160 --> 00:59:30.797
if you ever really describe it like this
but the way I' interpret it is it's it's

00:59:30.880 --> 00:59:33.757
like it's like this thing it's it's
actually a thing in philosophy it's like

00:59:33.840 --> 00:59:37.197
aesthetics are not just appearances
aesthetics go all the way to like deep

00:59:37.280 --> 00:59:41.837
underlining underlying meaning right it's
like I'm not a physicist one of the things

00:59:41.920 --> 00:59:44.397
I've heard physicists say is one of the
things you start to get a sense of when

00:59:44.480 --> 00:59:47.677
a theory might be correct is when it's
beautiful right like you know there

00:59:47.760 --> 00:59:50.957
right and so so so there's something
and you you feel the same thing by the way

00:59:51.040 --> 00:59:53.997
in like human psychology Right?
You know, when when you're

00:59:54.080 --> 00:59:57.037
experiencing awe, right? You know,
there's like a there's like a there's a

00:59:57.120 --> 00:59:59.677
simplicity to it when when you're having
an honest interaction with somebody,

00:59:59.760 --> 01:00:02.877
there's an aesthetic, I would say, calm
that comes over you cuz you're actually

01:00:02.960 --> 01:00:05.597
being fully honest and trying to hide
yourself, right? So there there so so

01:00:05.680 --> 01:00:08.397
it's like this very
deep sense of aesthetics

01:00:08.480 --> 01:00:10.317
>> and he would trust
that judgment that he

01:00:10.400 --> 01:00:12.957
had deep down like even even if the

01:00:13.040 --> 01:00:15.677
engineering teams are saying this is uh

01:00:15.760 --> 01:00:19.997
this is too difficult. Even if
the whatever the finance folks are saying

01:00:20.080 --> 01:00:23.357
this is ridiculous the uh the supply
chain all that kind of stuff this makes

01:00:23.440 --> 01:00:27.197
this impossible to m we can't do this
kind of material uh this has never been

01:00:27.280 --> 01:00:30.237
done before so on and so
forth he just sticks by it

01:00:30.320 --> 01:00:34.717
>> well I mean who makes a phone
out of aluminum right like [laughter]

01:00:34.800 --> 01:00:37.757
nobody else would have done that and now
of course if your phone was made out of

01:00:37.840 --> 01:00:41.037
aluminum what you know how crude what
a kind of caveman would you have to be to

01:00:41.120 --> 01:00:44.317
have a phone that's made out of plastic
like right so like so it's just this

01:00:44.400 --> 01:00:47.197
very right and you know look it's it's
there's a thousand different ways to

01:00:47.280 --> 01:00:49.437
look at this. But one of the things
is just like, look, these things are

01:00:49.520 --> 01:00:52.637
central to your lives. Like you're
with your phone more than you're with

01:00:52.720 --> 01:00:55.037
anything else. Like it's in your it's
going to be in your hand. I mean, he you

01:00:55.120 --> 01:00:57.277
know, you know this, he thought very
deeply about what it meant for something

01:00:57.360 --> 01:00:58.397
to be in your hand all day long.

01:00:58.480 --> 01:01:01.117
>> Yeah. >> Well,
for example, he a a here's an

01:01:01.200 --> 01:01:04.317
interesting design thing. Like he he
never wanted my understanding is he

01:01:04.400 --> 01:01:08.237
never wanted an iPhone to have a screen
larger than you could reach with your

01:01:08.320 --> 01:01:11.597
thumb one-handed. [clears throat] >>
And so he he was actually opposed to the

01:01:11.680 --> 01:01:14.077
idea of making the phones larger. And I
don't know if you have this experience

01:01:14.160 --> 01:01:17.037
today, but let's say there are certain
moments in your day when you might be

01:01:17.120 --> 01:01:20.237
like um only have one hand available um
and you might want to be on your phone.

01:01:20.320 --> 01:01:22.877
>> Yeah. [snorts] >> And you're
trying to like [laughter]

01:01:22.960 --> 01:01:25.517
send a text and you your thumb
can't reach the send button.

01:01:25.600 --> 01:01:28.077
>> Yeah. I mean there's pros and cons,
right? And then there's like folding

01:01:28.160 --> 01:01:32.237
phones which I would love to know what
he thought thinks about them. Uh but I

01:01:32.320 --> 01:01:35.277
mean is there something you could
also just linger on cuz he's one of the

01:01:35.360 --> 01:01:38.717
interesting um figures in the history of

01:01:38.800 --> 01:01:42.077
technology. What makes him what makes

01:01:42.160 --> 01:01:43.837
him as successful as he was? What makes

01:01:43.920 --> 01:01:46.077
him as interesting as he was? Uh what

01:01:46.160 --> 01:01:51.597
made him um so productive and important
in um in in in the development of

01:01:51.680 --> 01:01:54.157
technology. >> He had
an integrated worldview. So the

01:01:54.240 --> 01:01:57.997
the the properly designed device that
had the correct functionality that had

01:01:58.080 --> 01:02:02.877
the deepest understanding of the user
that was the most beautiful, right? Like

01:02:02.960 --> 01:02:04.877
it had to be all of those things, right?

01:02:04.960 --> 01:02:07.757
It was he basically would drive to as
close to perfect as you could possibly

01:02:07.840 --> 01:02:10.877
get, right? And I I you know I suspect
that he never quite you know thought he

01:02:10.960 --> 01:02:12.877
ever got there because most great
creators you know are generally

01:02:12.960 --> 01:02:15.997
dissatisfied. You know you read accounts
later on and all they can all they can

01:02:16.080 --> 01:02:18.477
see are the flaws in their creation.
But like he got as close to perfect each

01:02:18.560 --> 01:02:22.077
step of the way as he could possibly get
with the with the constraints of the of

01:02:22.160 --> 01:02:25.277
the technology of his time. Um and then
you know look he was you know sort of

01:02:25.360 --> 01:02:28.077
famous in the Apple model. It's like
look they they they will you know this

01:02:28.160 --> 01:02:31.757
this headset that they just came
out with like it's like a decade long

01:02:31.840 --> 01:02:34.717
project right? It's like and they're
just going to sit there and tune and

01:02:34.800 --> 01:02:37.757
tune and polish and polish and tune
and polish and tune and polish until it is

01:02:37.840 --> 01:02:39.837
as perfect as anybody could
possibly make anything.

01:02:39.920 --> 01:02:41.677
>> Yeah. >> And then
this goes to the the the way

01:02:41.760 --> 01:02:44.637
that people describe working with him
was which is you know there was a

01:02:44.720 --> 01:02:47.277
terrifying aspect of working with him
which is you know he was you know he was

01:02:47.360 --> 01:02:51.517
very tough. Um, but there was this thing
that everybody I've ever talked to who

01:02:51.600 --> 01:02:55.677
worked for him says that they all say
the following which is he we did the

01:02:55.760 --> 01:02:58.637
best work of our lives when we
worked for him because he set the bar

01:02:58.720 --> 01:03:01.437
incredibly high and then he supported us
with everything that he could to let us

01:03:01.520 --> 01:03:04.797
actually do work of that quality.
So a lot of people who were at Apple spend

01:03:04.880 --> 01:03:07.357
the rest of their lives trying to find
another experience where they feel like

01:03:07.440 --> 01:03:09.277
they're able to hit
that quality bar again.

01:03:09.360 --> 01:03:12.717
>> Even if it in retrospect
or during it felt like suffering.

01:03:12.800 --> 01:03:17.677
>> Yeah. Exactly. What does
that teach you about the human

01:03:17.760 --> 01:03:21.437
condition? Huh? >> So look,
so I say exactly. So the

01:03:21.520 --> 01:03:25.197
Silicon Valley I mean look he's not you
know George Patton in the you know in

01:03:25.280 --> 01:03:26.317
the army like you know there are many

01:03:26.400 --> 01:03:27.757
examples in other fields you know that

01:03:27.840 --> 01:03:32.317
are like this. Um uh uh

01:03:32.400 --> 01:03:35.597
specifically in in tech it's actually
I find it very interesting. There's the

01:03:35.680 --> 01:03:38.797
Apple way which is polish polish polish
and don't ship until it's as perfect as

01:03:38.880 --> 01:03:42.237
you can make it. And then there's the
sort of the other approach which is the

01:03:42.320 --> 01:03:45.837
sort of incremental hacker mentality
which basically says ship early and

01:03:45.920 --> 01:03:49.437
often and iterate. And one of the things
I find really interesting is I'm now 30

01:03:49.520 --> 01:03:53.277
years into this like there are
very successful companies on

01:03:53.360 --> 01:03:58.237
both sides of that approach right
um like that is a fundamental

01:03:58.320 --> 01:04:02.157
difference right in how to operate and
how to build and how to create that you

01:04:02.240 --> 01:04:06.397
have world-class companies operating
in both ways. Um, and I don't think the

01:04:06.480 --> 01:04:09.197
question of like which is the superior
model is anywhere close to being

01:04:09.280 --> 01:04:13.277
answered like and my suspicion is the
answer is do both. The answer is you

01:04:13.360 --> 01:04:17.117
actually want both. They lead to different
outcomes. Software tends to do

01:04:17.200 --> 01:04:19.117
better um with the iterative approach.

01:04:19.200 --> 01:04:23.837
Um hardware tends to do better with the uh
you know sort of wait and make it

01:04:23.920 --> 01:04:28.317
perfect approach but again you can find
examples in in in in both directions.

01:04:28.400 --> 01:04:30.637
>> Oh so the jury
is still out on that one.

01:04:30.720 --> 01:04:35.837
Uh so back to mosaic.
So what uh it was textbased?

01:04:35.920 --> 01:04:38.397
>> Yeah. >> Uh Tim Berners Lee.

01:04:38.480 --> 01:04:41.117
>> Well, there was the web which was text
based, but there were no I mean there

01:04:41.200 --> 01:04:44.107
was like three websites.
There was like no content. There were no

01:04:44.190 --> 01:04:46.797
[clears throat] users. Like it wasn't
like it wasn't like a catalytic. It

01:04:46.880 --> 01:04:49.197
hadn't and by the way it was all
because it was all text. There were no

01:04:49.280 --> 01:04:52.237
documents. There were no images. There
were no videos. There were no Right. So

01:04:52.320 --> 01:04:55.677
So it was it was And then if if in the
beginning if you had to be on a next

01:04:55.760 --> 01:04:59.117
cube, right, you need to had a next
cube both to publish and to consume. So

01:04:59.200 --> 01:05:01.037
>> So there there were
6,000 bucks. You said

01:05:01.120 --> 01:05:03.197
>> there were limitations.
Yeah. $6,000 PC.

01:05:03.280 --> 01:05:05.357
They did not they did not sell very many.

01:05:05.440 --> 01:05:09.277
>> But then there was also there was also
FTP and there was Usenet, right? And

01:05:09.360 --> 01:05:12.237
there was, you know, a dozen other
basically there's Waste, which was an

01:05:12.320 --> 01:05:15.037
early search thing. There was Gopher,
which was an early menu based

01:05:15.120 --> 01:05:17.917
information retrieval system. There
there were like a dozen different sort

01:05:18.000 --> 01:05:20.957
of scattered ways that people would get
to information on on the internet. And

01:05:21.040 --> 01:05:24.157
so the the mosaic idea was basically
bring those all together, make the whole

01:05:24.240 --> 01:05:27.277
thing graphical, make it easy to use,
make it basically bulletproof so that

01:05:27.360 --> 01:05:30.797
anybody can do it. And then again,
just on the luck side, it so happened that

01:05:30.880 --> 01:05:34.157
this was right at the moment when
graphics when the guey sort of actually

01:05:34.240 --> 01:05:37.437
took off and we're now also used
to the guey that we think it's been around

01:05:37.520 --> 01:05:41.677
forever, but it didn't really it, you
know, the Macintosh brought it out in '

01:05:41.760 --> 01:05:44.637
85, but they actually didn't sell very
many Macs in the 80s. [clears throat] It

01:05:44.720 --> 01:05:46.557
was not that successful of a product.

01:05:46.640 --> 01:05:49.037
Um, it really was you needed Windows 3.0

01:05:49.120 --> 01:05:52.877
on PCs and that hit in about 92. Um, and

01:05:52.960 --> 01:05:56.877
so and we did Mosaic 92,93. So that sort
of it was like right at the moment when

01:05:56.960 --> 01:06:01.837
you could imagine actually having
a graphical user interface to right at all

01:06:01.920 --> 01:06:03.117
much less one to the internet.

01:06:03.200 --> 01:06:07.677
>> How how old did Windows 3 sell?
So it was that the really big

01:06:07.760 --> 01:06:11.117
>> that was the big bang >> the big
operating graphical operating system.

01:06:11.200 --> 01:06:12.717
>> Well this is the classic.
Okay Microsoft

01:06:12.800 --> 01:06:16.317
was operating on the other. So Steve
Steve Apple was running on the polish

01:06:16.400 --> 01:06:18.957
until perfect. Microsoft famously ran
on the other model which is ship and

01:06:19.040 --> 01:06:22.237
iterate. And so in the old line in those
days was Microsoft right version three

01:06:22.320 --> 01:06:25.117
of every Microsoft product. That's the
That's the good one, right? And so there

01:06:25.200 --> 01:06:28.397
there are you can you can find online
Windows one, Windows 2, nobody used them.

01:06:28.480 --> 01:06:31.757
>> Yeah. >> Actually, the original Windows
the in the original Microsoft Windows, the

01:06:31.840 --> 01:06:33.497
windows were non-over overlapping.

01:06:33.580 --> 01:06:35.837
[snorts] >> Um and so you
had these very small very

01:06:35.920 --> 01:06:40.317
low resolution screens and then you
had literally um it just didn't work. It

01:06:40.400 --> 01:06:44.397
wasn't ready yet. >> Well, and Windows 95,
I think, was a pretty big leap also.

01:06:44.480 --> 01:06:47.997
>> That was a big leap, too. Yeah. So,
that was like bang bang. Um and then, of

01:06:48.080 --> 01:06:50.637
course, Steve and then and then,
you know, in the fullness of time, Steve

01:06:50.720 --> 01:06:53.037
came back. Then the Mac start took off
again. And that was the third bang. And

01:06:53.120 --> 01:06:54.957
then the iPhone was a fourth bang.

01:06:55.040 --> 01:06:57.597
>> Such exciting times.
>> And then we were off off to the races

01:06:57.680 --> 01:07:01.437
>> cuz nobody could have known
what would be created from that.

01:07:01.520 --> 01:07:04.397
>> Well, Windows 3.1
or 3.0, Windows 3.0 to

01:07:04.480 --> 01:07:09.277
the iPhone was only 15 years,
>> right? Like it that ramp was in

01:07:09.360 --> 01:07:12.317
retrospect at the time it felt like
it took forever, but that in histo in

01:07:12.400 --> 01:07:15.837
historical terms like that was a very
fast ramp from even a graphical computer

01:07:15.920 --> 01:07:20.077
at all on your desk to the iPhone.
It was 15 years. Did you did you have a

01:07:20.160 --> 01:07:22.477
sense of what the internet will be
as you're looking through the window of

01:07:22.560 --> 01:07:26.317
Mosaic like like what you like there's

01:07:26.400 --> 01:07:28.477
just a few web pages for now.

01:07:28.560 --> 01:07:33.597
>> So the thing I had early on was I
was keeping at the time what there's

01:07:33.680 --> 01:07:36.477
disputes over what was the first blog
but I I had one of them that at least is

01:07:36.560 --> 01:07:40.557
a is a is a uh is a pos possible um at
least a runner up in the competition. Um

01:07:40.640 --> 01:07:42.957
and it it was what was called
the what's new page.

01:07:43.040 --> 01:07:47.597
>> Um uh and it was it was it was a
hardwired and I had distribution unfair

01:07:47.680 --> 01:07:50.157
advantage. I was I wired I put it
right in the browser. [laughter]

01:07:50.240 --> 01:07:52.797
>> I put it in the browser and then I put
my resume in the browser which also was

01:07:52.880 --> 01:07:55.917
was >> hilarious. >> But um [laughter]

01:07:56.000 --> 01:07:59.597
but um I was I I was keeping the not

01:07:59.680 --> 01:08:02.157
many people get to get to do that. So um

01:08:02.240 --> 01:08:07.197
>> the uh [laughter] >>
good good call early days.

01:08:07.280 --> 01:08:09.037
>> Yes. >> It's so interesting.

01:08:09.120 --> 01:08:12.157
>> I'm looking for my about Oh,
Mark is looking for a job.

01:08:12.240 --> 01:08:16.877
>> Yeah, exactly. >> Um [laughter] so
um >> Wow. So the West New page, I would

01:08:16.960 --> 01:08:20.717
literally get up every morning and I
would or every afternoon um and I would

01:08:20.800 --> 01:08:24.557
basically if you wanted to launch
a website, you would email me um and I

01:08:24.640 --> 01:08:27.197
would list it on the West New page
and that was how people discovered the new

01:08:27.280 --> 01:08:30.397
websites as they were coming out.
And I remember cuz it was like one it

01:08:30.480 --> 01:08:34.017
literally went from it was like one
every couple days to like one every day

01:08:34.100 --> 01:08:37.197
[laughter] to like two
every day. Boom boom boom.

01:08:37.280 --> 01:08:39.117
And then >> so you're
doing so that that blog was

01:08:39.200 --> 01:08:43.037
kind of doing the directory thing.
So like what was the homepage? Uh, so the

01:08:43.120 --> 01:08:45.597
homepage was just basically trying
to explain even what this thing is that

01:08:45.680 --> 01:08:49.197
you're looking at, right? Basic
basically basic instructions. Um, but

01:08:49.280 --> 01:08:51.277
then there was a button there was a
button that said what's new and what

01:08:51.360 --> 01:08:54.237
most people did was they went to for
obvious reasons went to what's new.

01:08:54.320 --> 01:08:57.997
>> But like it it was so it was so
mind-blowing at that point just the

01:08:58.080 --> 01:09:00.637
basic idea and this was this was like
you know this is the basic idea of the

01:09:00.720 --> 01:09:03.997
internet but people could see it for the
first time. The basic idea was look, you

01:09:04.080 --> 01:09:06.317
know, some, you know, it's like
literally it's like an Indian restaurant

01:09:06.400 --> 01:09:09.997
in like Bristol, England has like put
their menu on the web and people were

01:09:10.080 --> 01:09:15.037
like, "Wow, >> cuz like that's
the first restaurant menu on the web."

01:09:15.120 --> 01:09:16.717
>> Yeah. >> And I don't
have to be in Bristol and I

01:09:16.800 --> 01:09:18.797
don't know if I'm ever going to go
to Bristol and I don't even like Indian

01:09:18.880 --> 01:09:22.397
food and like wow, >> right?
Um and it was like that uh the

01:09:22.480 --> 01:09:26.637
first web uh the first streaming video
thing was a uh it was it was in another

01:09:26.720 --> 01:09:29.037
England some Oxford or something. Um,

01:09:29.120 --> 01:09:33.997
some [snorts] guy uh put uh his coffee
pot up as the first uh streaming uh uh

01:09:34.080 --> 01:09:37.117
video thing and he put it on the web cuz
he literally it was the coffee pot down

01:09:37.200 --> 01:09:39.197
the hall. >> Yeah. >> And he
wanted to see when he needed to

01:09:39.280 --> 01:09:42.957
go refill it. Um but there were you
know there was a point when there were

01:09:43.040 --> 01:09:45.117
thousands of people like
watching that coffee pot

01:09:45.200 --> 01:09:47.277
>> cuz it was the first
thing you could watch.

01:09:47.360 --> 01:09:48.477
>> Well, [laughter] but Right.

01:09:48.560 --> 01:09:52.877
>> Isn't [snorts] uh were you able
to kind of infer you know if that Indian

01:09:52.960 --> 01:09:56.797
restaurant could go online >>
then you're like >> they all will.

01:09:56.880 --> 01:09:58.317
>> They all will. >> Yeah. Exactly.

01:09:58.400 --> 01:10:00.957
>> So, you felt that >> Yeah.
Now, you know, look, it's still a

01:10:01.040 --> 01:10:03.117
stretch, right? It's still a stretch
because it's just like, okay, is it, you

01:10:03.200 --> 01:10:05.357
know, you're still in this zone which
is like, okay, is this a nerd thing? Is

01:10:05.440 --> 01:10:08.317
this a real person thing? >> Um,
by the way, we, you know, there was

01:10:08.400 --> 01:10:09.997
a wall of skepticism from the media.

01:10:10.080 --> 01:10:12.557
Like, they just like everybody was just
like, yeah, this is the crazy, is this

01:10:12.640 --> 01:10:15.197
just like dumb? This is not, you know,
this is not for regular people at that

01:10:15.280 --> 01:10:17.517
time. >> Um, and so
you had to think through

01:10:17.600 --> 01:10:20.957
that. And then, look, it was still it
was still hard to get on the internet at

01:10:21.040 --> 01:10:24.077
that point, right? So you could get kind
of this weird bastardized version if you

01:10:24.160 --> 01:10:26.237
were on AOL, which wasn't really real.

01:10:26.320 --> 01:10:31.277
Or you had to go like learn what an ISP
was. Um, you know, in those days PCs

01:10:31.360 --> 01:10:34.237
actually didn't have TCP IP drivers come
pre-installed. So you had to learn what

01:10:34.320 --> 01:10:37.917
a TCP IP driver was. You had to buy
a modem. You had to install driver

01:10:38.000 --> 01:10:42.397
software. Um, I have a comedy routine I
do some like 20 minutes long describing

01:10:42.480 --> 01:10:44.397
all the steps required to actually
get on the internet at this [laughter]

01:10:44.480 --> 01:10:48.637
point. Um and so you you had to you had
to look through these practical well and

01:10:48.720 --> 01:10:51.597
then and then uh
and then speed performance

01:10:51.680 --> 01:10:56.397
>> 144 modems right like it was like
watching you know glue dry um like and

01:10:56.480 --> 01:10:59.117
so you had to you had to there were
basically a sequence of bets that we

01:10:59.200 --> 01:11:01.517
made where you basically needed to look
through that current state of affairs

01:11:01.600 --> 01:11:04.797
and say actually there's going to be so
much demand for once people figure this

01:11:04.880 --> 01:11:06.957
out there's going to be so much demand
for it that all of these practical

01:11:07.040 --> 01:11:08.877
problems are are going to get fixed.

01:11:08.960 --> 01:11:15.117
Some people say that the anticipation
makes the the destination that much more

01:11:15.200 --> 01:11:17.437
exciting. >> Do you
remember progressive JPEGs?

01:11:17.520 --> 01:11:22.237
>> Yeah. Do I do I >> So
for for kids in the audience, right?

01:11:22.320 --> 01:11:24.797
>> For kids in the audience.
>> We you used to have to watch an image

01:11:24.880 --> 01:11:28.237
load like a line at a time, but it
turns out there was this thing with JPEGs

01:11:28.320 --> 01:11:30.717
where you could you could
load basically every fourth

01:11:30.800 --> 01:11:32.717
>> you could load like
every fourth uh line

01:11:32.800 --> 01:11:35.597
and then and then you could sweep back
through again. And so you could like

01:11:35.680 --> 01:11:38.797
render a fuzzy version image up front
and then it would like resolve into the

01:11:38.880 --> 01:11:41.917
detailed one. And that was like a big
UI breakthrough because it gave you

01:11:42.000 --> 01:11:45.677
something to watch. >> Yeah.
And uh you know there's

01:11:45.760 --> 01:11:50.557
applications in various
domains for that. Uh [laughter]

01:11:50.640 --> 01:11:52.557
>> well there's a big fight. There was
a big fight early on about whether there

01:11:52.640 --> 01:11:54.557
should be images on the web.
Um [clears throat]

01:11:54.640 --> 01:11:56.477
>> for that reason for like sexualization.

01:11:56.560 --> 01:11:58.397
>> No not not explicitly
that that did come

01:11:58.480 --> 01:12:01.277
up but it wasn't even that. It was more
just like all the serious. The argument

01:12:01.360 --> 01:12:04.157
went the purists basically said all
the serious information in the world is

01:12:04.240 --> 01:12:07.597
text. If you introduce images you you
basically going to bring in all the

01:12:07.680 --> 01:12:10.717
trivial stuff. You're going to bring
in magazines and you know all this crazy

01:12:10.800 --> 01:12:13.517
just you know stuff that you know people
you know it's going to distract from

01:12:13.600 --> 01:12:16.077
it's going to go take take the way
from being serious to being frivolous.

01:12:16.160 --> 01:12:20.717
>> Well was there any uh doomer
type arguments about uh the internet

01:12:20.800 --> 01:12:26.157
destroying all of human civilization
or destroying some fundamental fabric of

01:12:26.240 --> 01:12:29.037
human civilization. >> Yeah.
So those days it was all around

01:12:29.120 --> 01:12:33.277
crime and terrorism. Um so those
arguments happened. Um you know but

01:12:33.360 --> 01:12:35.757
there was no sense yet of the internet
having like an effect on politics

01:12:35.840 --> 01:12:38.957
because that was that was way too too
far off. But um there was an enormous

01:12:39.040 --> 01:12:41.197
panic at the time around cyber crime.

01:12:41.280 --> 01:12:43.917
There was like enormous panic that like
your credit card number would get stolen

01:12:44.000 --> 01:12:46.957
and you'd your life savings would be
drained and then you know criminals were

01:12:47.040 --> 01:12:50.557
going to there was oh um when we started
one of the things we did one of the the

01:12:50.640 --> 01:12:54.157
Netscape browser was the first widely
used piece of consumer software that had

01:12:54.240 --> 01:12:57.997
strong encryption built in. made it
available to ordinary people and at that

01:12:58.080 --> 01:13:02.317
time strong encryption was actually
illegal to export out of the US. So we

01:13:02.400 --> 01:13:05.597
could field that product in the US.
We could not export it because it was it

01:13:05.680 --> 01:13:09.197
was classified as ammunition. Um so
the Netscape browser was on a restricted

01:13:09.280 --> 01:13:11.997
list along with the tomahawk missile
as [snorts] being something that could not

01:13:12.080 --> 01:13:15.117
be exported. So we we had to make
a second version with deliberately weak

01:13:15.200 --> 01:13:19.197
encryption to sell overseas with a big
logo on the box saying do not trust this

01:13:19.280 --> 01:13:22.317
which it turns out makes it hard to sell
software uh when [laughter] it's got a

01:13:22.400 --> 01:13:25.917
big logo that says don't trust it.
Um, and then we had to spend five years

01:13:26.000 --> 01:13:30.237
fighting the US government to get them
to basically stop trying to do this. But

01:13:30.320 --> 01:13:33.197
because the fear the fear was terrorists
are going to use encryption, right, to

01:13:33.280 --> 01:13:36.157
like plot, you know, all these
all these all these things.

01:13:36.240 --> 01:13:39.357
>> Um, and then, you know, we we responded
with, well, actually, we need encryption

01:13:39.440 --> 01:13:42.317
to be able to secure systems so that the
terrorists and the criminals can't get

01:13:42.400 --> 01:13:45.517
into them. So that anyway,
that was the that was the 1990s fight.

01:13:45.600 --> 01:13:49.117
>> So, uh, can you say something about
some of the details of the software

01:13:49.200 --> 01:13:53.837
engineering challenges required to build
these browsers? I mean the engineering

01:13:53.920 --> 01:13:55.517
challenges of creating a product that

01:13:55.600 --> 01:13:58.157
hasn't really existed before that can

01:13:58.240 --> 01:14:02.077
have such uh almost like limitless uh

01:14:02.160 --> 01:14:04.317
impact on the world with the internet.

01:14:04.400 --> 01:14:06.877
>> So there was a really key bet
that we made at the time which is very

01:14:06.960 --> 01:14:10.157
controversial which was core to core
to how it was engineered which was are we

01:14:10.240 --> 01:14:13.197
optimizing for performance
um or for ease of creation.

01:14:13.280 --> 01:14:15.037
>> Yeah. And in those
days the pressure was

01:14:15.120 --> 01:14:17.997
very intense to optimize for performance
because the network connections were so

01:14:18.080 --> 01:14:19.917
slow and also the computers were so

01:14:20.000 --> 01:14:25.107
slow. Um and so if you had I mentioned
the progressive JPEGs like if if if

01:14:25.190 --> 01:14:26.957
[laughter] there there's
an alternate world in

01:14:27.040 --> 01:14:29.677
which we optimized for performance
and it just you had just a much more

01:14:29.760 --> 01:14:34.317
pleasant experience right up front. But
what we got by not doing that was we got

01:14:34.400 --> 01:14:38.317
ease of creation. And the way that we
got ease of creation was all of the

01:14:38.400 --> 01:14:41.917
protocols and formats were
in text not in binary. Mhm.

01:14:42.000 --> 01:14:44.797
>> Um, [clears throat] and so HTTP
isn't text, by the way. And this was an

01:14:44.880 --> 01:14:47.757
internet tradition, by the way, that we
picked up, but we continued it. HTTP is

01:14:47.840 --> 01:14:51.117
text. Um, and HTML is text,
and then every else, everything else that

01:14:51.200 --> 01:14:54.957
followed is text. Um, as a result,
and by the way, you can imagine purist

01:14:55.040 --> 01:14:57.917
engineers saying, "This is insane. You
have very limited bandwidth. Why are you

01:14:58.000 --> 01:15:00.477
wasting any time sending text?
You should be encoding the stuff into

01:15:00.560 --> 01:15:03.677
binary, and it'll be much faster." And
of course, the answer is that's correct.

01:15:03.760 --> 01:15:06.717
Um, but what you get when you make it
text is all of a sudden, well, the big

01:15:06.800 --> 01:15:09.277
breakthrough was the view source
function, right? So the fact that you

01:15:09.360 --> 01:15:12.477
could look at a web page, you could hit
view source and you could see the HTML,

01:15:12.560 --> 01:15:15.037
that was how people learned
how to make web pages, right?

01:15:15.120 --> 01:15:19.117
>> It's so interesting because
the stuff we take for granted now

01:15:19.200 --> 01:15:22.957
is uh man that was fundamental to the
development of the web to be able to

01:15:23.040 --> 01:15:25.197
have HTML just right there. All the

01:15:25.280 --> 01:15:30.317
ghetto mess that is HTML, all [snorts]
the sort of almost biological like

01:15:30.400 --> 01:15:34.877
messiness of HTML and then
having the browser try

01:15:34.960 --> 01:15:38.237
to interpret that mess. >> Yeah,
exactly. to show something reasonable.

01:15:38.320 --> 01:15:40.877
>> Well, and then there was this internet
principle that we inherited which was

01:15:40.960 --> 01:15:45.117
emit, what was it? Emit cautiously, emit
conservatively, interpret liberally. So,

01:15:45.200 --> 01:15:48.397
it basically meant if you're the design
principle was if you're if you're

01:15:48.480 --> 01:15:51.437
creating like a web editor that's going
to emit HTML, like do it as cleanly as

01:15:51.520 --> 01:15:53.597
you can, >> but you
actually want the browser to

01:15:53.680 --> 01:15:56.477
interpret liberally, which is you
actually want users to be able to make

01:15:56.560 --> 01:15:58.557
all kinds of mistakes
and for it to still work.

01:15:58.640 --> 01:16:01.997
>> Yeah. And so the browser rendering
engines to this day have all of this

01:16:02.080 --> 01:16:05.277
spaghetti code crazy stuff where they
can they're they're resilient to all

01:16:05.360 --> 01:16:08.557
kinds of crazy HTML mistakes. And so
and literally what I always had in my head

01:16:08.640 --> 01:16:10.957
is like there's an 8-year-old or an
11-year-old somewhere and they're doing

01:16:11.040 --> 01:16:13.117
a view source. They're doing a cut
and paste and they're trying to make a web

01:16:13.200 --> 01:16:16.797
page for their turtle or whatever >>
and like they leave out a slash and they

01:16:16.880 --> 01:16:19.197
leave out an angle bracket and they
do this and they do that and it still

01:16:19.280 --> 01:16:21.837
works. It's it's it's also like I don't

01:16:21.920 --> 01:16:23.917
often think about this but you know

01:16:24.000 --> 01:16:26.797
programming you know C++ C++ all those

01:16:26.880 --> 01:16:31.117
languages list the compiled languages
the interpreted languages Python Pearl

01:16:31.200 --> 01:16:35.277
all that they the brace have to be all
correct it's like everything has to be

01:16:35.360 --> 01:16:38.157
perfect >> brutal >> and then >> autistic

01:16:38.240 --> 01:16:41.917
>> you forget all right
it's systematic and

01:16:42.000 --> 01:16:43.997
rigorous let's go there [laughter]

01:16:44.080 --> 01:16:48.317
but you forget that the uh uh the web

01:16:48.400 --> 01:16:50.637
with JavaScript script eventually.

01:16:50.720 --> 01:16:54.157
>> Uh, and HTML is allowed to be messy in

01:16:54.240 --> 01:16:57.437
the way for the first time

01:16:57.520 --> 01:17:01.517
>> messy in the way biological systems
could be messy. It's like the only thing

01:17:01.600 --> 01:17:04.557
computers were allowed to be
messy on for the first time.

01:17:04.640 --> 01:17:08.317
>> It used to offend me. So I I I grew
up on Unix. So I I I I worked on Unix. I

01:17:08.400 --> 01:17:11.277
was a Unix native for all the way
through this period. Um, and so and it

01:17:11.360 --> 01:17:14.477
used to drive me bananas when it would
do the the segmentation fault in the

01:17:14.560 --> 01:17:17.597
core dump file. Just like it's like,
you know, it's like literally there's like

01:17:17.680 --> 01:17:21.117
an error in the code. the math
is off by one and it core >> dumps

01:17:21.200 --> 01:17:23.517
>> and I'm in the core dump trying
to analyze it and trying to reconstruct

01:17:23.600 --> 01:17:26.237
what and I'm just like this is ridiculous
like the computer ought to be

01:17:26.320 --> 01:17:29.357
smart enough to be able to know that if
it's off by one okay fine and it keeps

01:17:29.440 --> 01:17:31.757
running >> and I would go
ask all the experts like

01:17:31.840 --> 01:17:33.837
why can't it just keep running
and they'd explain to me well because all

01:17:33.920 --> 01:17:37.357
the downstream repercussions and blah
blah and I'm like this still like you

01:17:37.440 --> 01:17:40.317
know this is we're forcing the human

01:17:40.400 --> 01:17:45.037
creator to live to your point in this
hyperlit literal world of perfection.

01:17:45.120 --> 01:17:48.077
>> Yeah. And I [snorts] was just like
that's that's just that's just bad. And

01:17:48.160 --> 01:17:50.557
by the way, you know, what happens with
that, of course, just what what happened

01:17:50.640 --> 01:17:53.837
with with coding at that point, which
is you get a high priesthood. You know,

01:17:53.920 --> 01:17:57.037
there's a small number of people who are
really good at doing exactly that. Most

01:17:57.120 --> 01:17:59.677
people can't and most people are
excluded from it. And so, actually, that

01:17:59.760 --> 01:18:03.437
that was where that there's where I
picked up that idea was um uh was like

01:18:03.520 --> 01:18:06.477
no, no, you want you want you want these
things to be resilient to error in all

01:18:06.560 --> 01:18:09.437
kinds. And this this would drive
the purists absolutely crazy. Like I got

01:18:09.520 --> 01:18:12.477
attacked on this like a lot cuz yeah I
mean like every time I you know all the

01:18:12.560 --> 01:18:15.517
purists who were like into all this like
markup language stuff and formats and

01:18:15.600 --> 01:18:18.077
codes and all this stuff they would be
like you know you can't you're you're

01:18:18.160 --> 01:18:21.197
encouraging bad behavior because >>
Oh so they wanted the browser to give

01:18:21.280 --> 01:18:26.077
you an a sec fault error anytime there was
a >> Yeah. Yeah. They wanted to be a cop

01:18:26.160 --> 01:18:30.157
right. They wanted Yeah. That that
was a very any any properly training

01:18:30.240 --> 01:18:33.277
credential engineer [laughter] would
be like that's not how you build these

01:18:33.360 --> 01:18:36.157
systems. >> That's such a bold move
to say no it doesn't have to be.

01:18:36.240 --> 01:18:38.477
>> Yeah. Now, like I said, the the good
news for me is the internet kind of had

01:18:38.560 --> 01:18:42.237
that tradition already. Um, but we
but having said that, like we pushed it,

01:18:42.320 --> 01:18:44.237
>> we pushed it way out. But the other
thing we did, going back to the

01:18:44.320 --> 01:18:46.637
performance thing, was we gave up
a lot of performance. We made that that

01:18:46.720 --> 01:18:49.517
initial experience for the first few
years was pretty painful. But but the

01:18:49.600 --> 01:18:52.877
bet there was actually an economic bet
which was basically the demand for the

01:18:52.960 --> 01:18:56.077
web would basically mean that there
would be a surge in supply of broadband.

01:18:56.160 --> 01:19:00.237
Like we we because the question was okay
how do you get how do you how do you get

01:19:00.320 --> 01:19:04.797
the phone companies which are not famous
in those days for doing new things at

01:19:04.880 --> 01:19:06.637
huge cost for like speculative reasons.

01:19:06.720 --> 01:19:09.597
Like how do you get them to build out
broadband you know spend billions of

01:19:09.680 --> 01:19:12.317
dollars doing that and you know you
could go meet with them and try to talk

01:19:12.400 --> 01:19:15.197
them into it or you could just have
a thing where it's just very clear that

01:19:15.280 --> 01:19:17.677
it's going to be that they that people
love that's going to be better

01:19:17.760 --> 01:19:21.197
>> if it's faster. And so that that there
there was a period there and this was

01:19:21.280 --> 01:19:23.597
this was fraught with some peril but there
was a period there where it's like

01:19:23.680 --> 01:19:27.757
we knew the experience was suboptimized
because we were trying to force the

01:19:27.840 --> 01:19:29.677
emergence of demand for broadband.

01:19:29.760 --> 01:19:31.997
>> Sure. >> Which
is in fact what happened.

01:19:32.080 --> 01:19:36.557
>> So you had to figure out how
to display this text HTML text.

01:19:36.640 --> 01:19:40.317
>> So the blue links and the purple
links and there's no standards. Is there

01:19:40.400 --> 01:19:44.877
standards at that time? >> No.
So >> there really still isn't.

01:19:44.960 --> 01:19:48.797
Well, there's like there's
implied implied standards, right?

01:19:48.880 --> 01:19:51.517
>> And they, you know, there's all these
kinds of new features that are being

01:19:51.600 --> 01:19:55.437
added with like CSS, what like what
kind of stuff a browser should be able to

01:19:55.520 --> 01:19:57.117
support, features within languages

01:19:57.200 --> 01:20:03.197
within JavaScript and so on.
>> But you you b you're setting standards

01:20:03.280 --> 01:20:05.357
on the fly. >> Yeah. >> Yourself.

01:20:05.440 --> 01:20:07.837
>> Yeah. >> Well, to this day,
if you if you create

01:20:07.920 --> 01:20:11.597
a web page that has no CSS stylesheet,
the browser will render it however it

01:20:11.680 --> 01:20:14.317
wants to. >> Yeah. >> Right.
So this was one of the things

01:20:14.400 --> 01:20:17.597
there was this idea this idea at the
time in how these systems were built

01:20:17.680 --> 01:20:21.677
which is separation of content from
format or separation of uh yeah content

01:20:21.760 --> 01:20:25.037
from appearance.
>> Um and that's still people don't really

01:20:25.120 --> 01:20:27.597
use that anymore because everybody wants
to determine how things look and so they

01:20:27.680 --> 01:20:30.077
use CSS but um it's still in there that

01:20:30.160 --> 01:20:31.277
you can just let the browser do all the

01:20:31.360 --> 01:20:35.037
work. I still like the like uh really

01:20:35.120 --> 01:20:37.597
basic websites, but that could be just

01:20:37.680 --> 01:20:42.877
old school kids these days with their
fancy responsive websites that don't

01:20:42.960 --> 01:20:45.437
actually have much content
but have a lot of visual elements.

01:20:45.520 --> 01:20:48.797
>> Well, that's one of the things that's
fun about chat, you know, about JGBT is

01:20:48.880 --> 01:20:51.357
like >> back to the basics.
>> Back [clears throat] to just text.

01:20:51.440 --> 01:20:53.837
>> Yeah. >> Right. And it,
you know, there is this

01:20:53.920 --> 01:20:58.397
pattern in human creativity and media
where you end up back at text. And I

01:20:58.480 --> 01:21:01.117
think there's, you know, there's
something powerful in there. Is there

01:21:01.200 --> 01:21:04.077
some other stuff you remember like
the purple links? There were some

01:21:04.160 --> 01:21:05.757
interesting design decisions that had to

01:21:05.840 --> 01:21:11.197
kind of come up that uh we have today or
we don't have today that were temporary.

01:21:11.280 --> 01:21:15.517
>> So uh we made I made the background
gray. I hated reading text on white uh

01:21:15.600 --> 01:21:17.597
uh backgrounds and so I made
the background gray. Everybody can.

01:21:17.680 --> 01:21:19.517
>> Do you regret? >> No.
Do you regret this?

01:21:19.600 --> 01:21:21.517
>> No. No. No.
That's that decision I think

01:21:21.600 --> 01:21:24.717
has been reversed. But but now I'm
happy though because now dark mode is the

01:21:24.800 --> 01:21:27.997
thing. So >> So it wasn't
about gray. It was just you

01:21:28.080 --> 01:21:29.197
didn't want white background.

01:21:29.280 --> 01:21:32.317
>> Strain my eyes. Stranger eyes.

01:21:32.400 --> 01:21:36.637
Interesting. Um, and then there's
a bunch of other decisions. I'm sure

01:21:36.720 --> 01:21:40.077
there's an interesting history
of the development of HTML and CSS and how

01:21:40.160 --> 01:21:42.957
those interface and JavaScript. And

01:21:43.040 --> 01:21:45.837
there's this whole Java applet thing.

01:21:45.920 --> 01:21:47.997
>> Well, the big one probably JavaScript.

01:21:48.080 --> 01:21:51.837
CSS was after me, so I didn't that was
not me. But, um, JavaScript was the big

01:21:51.920 --> 01:21:55.757
JavaScript maybe was the biggest of the
whole thing. That was us. Um, and um,

01:21:55.840 --> 01:21:59.197
and that was basically a bet. It was a bet
on two things. One is that the world

01:21:59.280 --> 01:22:01.437
wanted a new front-end scripting language.

01:22:01.520 --> 01:22:04.237
>> Um, and then the other was we I
thought at the time the world wanted a new

01:22:04.320 --> 01:22:07.437
backend scripting language. Um,
so JavaScript was designed from the

01:22:07.520 --> 01:22:11.277
beginning to be both front end and back
end. And then it failed as a backend

01:22:11.360 --> 01:22:15.357
scripting language and uh Java 1 um
for a long time and then Python, Pearl and

01:22:15.440 --> 01:22:20.317
other things, PHP um and Ruby.
But now JavaScript is back. And so

01:22:20.400 --> 01:22:23.837
>> I wonder if everything in the end
will run on JavaScript. It it see it seems

01:22:23.920 --> 01:22:28.157
like it is the um and by the way,
let me give a shout out to uh to um uh uh

01:22:28.240 --> 01:22:32.557
Brendan Ike uh was the uh basically
the oneman inventor of um of JavaScript.

01:22:32.640 --> 01:22:33.837
>> If you're interested
to learn more about

01:22:33.920 --> 01:22:37.677
Brendan Ike, he's been on this podcast
previously. [laughter] >> Exactly.

01:22:37.760 --> 01:22:39.277
>> So, he wrote JavaScript over a summer.

01:22:39.360 --> 01:22:42.637
Um and it it I I mean I think it is fair
it is fair to say now that it's the most

01:22:42.720 --> 01:22:46.717
widely used language in the world and it
seems to only be gaining in in um in its

01:22:46.800 --> 01:22:50.157
uh in its range of adoption.
In the software world, there's quite a few

01:22:50.240 --> 01:22:52.077
stories of somebody over a weekend or

01:22:52.160 --> 01:22:55.197
over a week or over a summer writing

01:22:55.280 --> 01:22:59.837
some of the most uh impactful,
revolutionary pieces of software

01:22:59.920 --> 01:23:01.917
>> ever. That that should be inspiring.

01:23:02.000 --> 01:23:03.437
Yes, >> very inspiring.
I'll give you another

01:23:03.520 --> 01:23:07.837
one. SSL. Um, so SSL was the security
protocol. That was us. And that was a

01:23:07.920 --> 01:23:10.797
crazy idea at the time, which was let's
take all the native protocols and let's

01:23:10.880 --> 01:23:13.837
wrap them in a security wrapper.
That was a guy named Kip Hickman who wrote

01:23:13.920 --> 01:23:18.477
that over a summer. Uh, one guy. Um and
then look today sitting here today like

01:23:18.560 --> 01:23:22.317
the transformer like at Google was a
small handful of people and then you

01:23:22.400 --> 01:23:26.637
know the number of people who have did
like the core work on GPT it's not that

01:23:26.720 --> 01:23:31.197
many people it's a pretty small handful
of people um and so yeah the pattern in

01:23:31.280 --> 01:23:35.597
software repeatedly over a very long
time has been it's it's a Jeff Jeff

01:23:35.680 --> 01:23:39.437
Bezos always had the two pizza rule uh
for teams at Amazon which is any team

01:23:39.520 --> 01:23:42.717
needs to be able to be fed with two
pizzas if you need the third pizza you

01:23:42.800 --> 01:23:45.997
have too many people and I I think
that's I think that's I I think it's

01:23:46.080 --> 01:23:47.677
actually the one pizza rule. >> Yeah.

01:23:47.760 --> 01:23:51.677
>> For the for the really creative work,
I think it's two people, three people.

01:23:51.760 --> 01:23:55.117
>> Well, that's you see that with certain
open source projects, like so much is

01:23:55.200 --> 01:23:56.397
done by like one or two people.

01:23:56.480 --> 01:24:00.397
>> Yeah. like it's it's it's so incredible
and that's why you see that gives me so

01:24:00.480 --> 01:24:02.877
much hope about the open source movement

01:24:02.960 --> 01:24:06.557
in this new age of AI where um you know

01:24:06.640 --> 01:24:10.397
just recently having had a conversation
with with Mark Zuckerberg of all people

01:24:10.480 --> 01:24:13.037
who's all in on open source which is so

01:24:13.120 --> 01:24:16.157
interesting to see and so inspiring to

01:24:16.240 --> 01:24:18.957
see cuz like releasing these models it

01:24:19.040 --> 01:24:21.037
is scary it is potentially very

01:24:21.120 --> 01:24:23.517
dangerous and we'll talk about that but

01:24:23.600 --> 01:24:26.317
it's also like if you believe in the

01:24:26.400 --> 01:24:31.597
goodness of most people and in the skill
set of most people and the desire to do

01:24:31.680 --> 01:24:35.757
good in the world. That's really
exciting cuz it's not putting it these

01:24:35.840 --> 01:24:39.277
models into the centralized control
of big corporations, the government and so

01:24:39.360 --> 01:24:41.117
on. It's putting it in the in the hands

01:24:41.200 --> 01:24:46.717
of a teen teenage kid with like a dream
in his eyes. I don't know. That's um

01:24:46.800 --> 01:24:49.677
>> that's beautiful. >> And look
this stuff AI ought to make the

01:24:49.760 --> 01:24:53.277
individual coder obviously far more
productive, right? By like, you know, a

01:24:53.360 --> 01:24:57.597
thousandx or something. And so you ought
to open source like the not just the

01:24:57.680 --> 01:25:00.717
future of open source AI but the future
of open source everything. We ought to

01:25:00.800 --> 01:25:04.557
have a world now of super coders right
who are building things as open source

01:25:04.640 --> 01:25:08.637
with one or two people that were
inconceivable you know 5 years ago. Um

01:25:08.720 --> 01:25:11.037
you know the level of kind of hyper
productivity we're going to get out of

01:25:11.120 --> 01:25:12.797
our best and brightest I
think is going to go way up.

01:25:12.880 --> 01:25:15.917
>> It's going to be interesting. We'll
we'll talk about it. But let's just to

01:25:16.000 --> 01:25:18.557
linger a little bit on Netscape.

01:25:18.640 --> 01:25:22.637
Netscape was acquired in 1999 for 4.3

01:25:22.720 --> 01:25:26.957
billion by AOL. What was that? Uh what
was that like? What was what were some

01:25:27.040 --> 01:25:30.477
memorable aspects of that?
>> Well, that was the height of the dot

01:25:30.560 --> 01:25:35.197
boom bubble bust. I mean, that was the
that was the frenzy. Um if you watch

01:25:35.280 --> 01:25:38.317
Succession, that was the that was like
what they did in the fourth season with

01:25:38.400 --> 01:25:41.837
uh the with Gojo and the merger with the
with their so it was like the height of

01:25:41.920 --> 01:25:45.437
like one of those kind of dynamics. And so
>> would you recommend Succession by the

01:25:45.520 --> 01:25:47.617
way? I'm more of a Yellowstone guy.

01:25:47.700 --> 01:25:51.917
[laughter] Yellowstone's very American.
I'm I'm very proud of you. That's that is

01:25:52.000 --> 01:25:55.437
>> I just talked to Matthew McConna
and I'm full on Texan at this point.

01:25:55.520 --> 01:25:57.437
>> Good. I hardily approve. Um

01:25:57.520 --> 01:26:02.717
>> and uh he will be doing the sequel
to Yellowstone. So very exciting. Anyway,

01:26:02.800 --> 01:26:06.797
>> I can't wait. >> Uh so
that's a rude interruption by me

01:26:06.880 --> 01:26:09.597
uh by way of succession.

01:26:09.680 --> 01:26:12.557
Uh, so that was at the height of the

01:26:12.640 --> 01:26:16.957
>> deal making and money and just the fur
flying and like craziness. And so yeah,

01:26:17.040 --> 01:26:19.837
it was just one of those it was just
like I mean this the entire Netscape

01:26:19.920 --> 01:26:23.437
thing from start to finish was four years.
Um, which was like for for one of

01:26:23.520 --> 01:26:26.237
these companies it's just like
incredibly fast you know it we went

01:26:26.320 --> 01:26:29.437
public 18 months after we got after
we were founded which virtually never

01:26:29.520 --> 01:26:33.117
happens. So it was just this incredibly
fast kind of meteor streaking across the

01:26:33.200 --> 01:26:36.557
sky. Um and then of course it was this
and then there was just this explosion

01:26:36.640 --> 01:26:39.517
right that happened because then it
was almost immediately followed by the do

01:26:39.600 --> 01:26:42.637
crash. >> It was then
followed by AOL buying Time

01:26:42.720 --> 01:26:46.077
Warner which again is the succession
guys kind of play with that uh which

01:26:46.160 --> 01:26:49.437
turned out to be a disastrous deal. Um
you know one of the famous you know kind

01:26:49.520 --> 01:26:53.597
of disasters in business history. Um and
then um and then you know what became an

01:26:53.680 --> 01:26:56.717
internet depression on the other side of
that. But then in that depression in the

01:26:56.800 --> 01:27:01.517
2000s was the beginning of broadband
and smartphones and web 2.0 know right and

01:27:01.600 --> 01:27:05.597
then social media and search and every SAS
and everything that came out of that. So

01:27:05.680 --> 01:27:10.317
>> what did you learn from just the the
acquisition? I mean this is so much money.

01:27:10.400 --> 01:27:12.557
>> What you what what's interesting cuz I

01:27:12.640 --> 01:27:15.037
it must have been very new to you that

01:27:15.120 --> 01:27:19.197
the software stuff you can make so much

01:27:19.280 --> 01:27:22.717
money. There's so much money swimming
around. I mean I'm sure the ideas of

01:27:22.800 --> 01:27:24.557
investment were starting
to get born there.

01:27:24.640 --> 01:27:26.237
>> Yes. Let me get so
let me lay it lay it.

01:27:26.320 --> 01:27:28.637
So here's here's a thing I I don't know
if I figured out then but figured out

01:27:28.720 --> 01:27:32.797
later which is um software is a technology
that it's like a you know the

01:27:32.880 --> 01:27:36.317
concept of the philosopher stone the
philosopher stone and alchemy transmutes

01:27:36.400 --> 01:27:38.877
light into gold and Newton spent 20
years trying to find the philosopher

01:27:38.960 --> 01:27:42.397
stone never got there. Nobody's ever
figured it out. Software is our modern

01:27:42.480 --> 01:27:45.277
philosopher stone. And in economic uh

01:27:45.360 --> 01:27:50.477
terms it transmutes labor into capital >>
which is like a super interesting thing.

01:27:50.560 --> 01:27:52.797
And by the way, like Karl Marx
is rolling over in his grave right now

01:27:52.880 --> 01:27:56.637
because of course that's complete
reputation of his entire theory. Um

01:27:56.720 --> 01:28:00.717
transduc labor into capital which
is which is as follows is somebody sits

01:28:00.800 --> 01:28:03.837
down at a keyboard
and types a bunch of stuff in

01:28:03.920 --> 01:28:06.317
>> and [clears throat] a capital
asset comes out the other side and then

01:28:06.400 --> 01:28:11.117
somebody buys that capital asset for
a billion dollars. Like that's amazing,

01:28:11.200 --> 01:28:14.397
right? It's literally creating value
right out of thin air, right? Out of out

01:28:14.480 --> 01:28:19.037
of purely human thought, right? Um,
and so that that that's there are many

01:28:19.120 --> 01:28:21.997
things that make software magical
and special, but that's the economics.

01:28:22.080 --> 01:28:23.997
>> I wonder what Markx would
have thought about that.

01:28:24.080 --> 01:28:26.637
>> Oh, he would have completely broke
his brain because of course the whole the

01:28:26.720 --> 01:28:30.237
whole thing was he was you could he
you know that kind of technology is

01:28:30.320 --> 01:28:33.517
inconceivable when he was alive. It was
all it was all industrial era stuff and

01:28:33.600 --> 01:28:37.517
so the any kind of machinery necessarily
involved huge amounts of capital and

01:28:37.600 --> 01:28:40.317
then labor was on the on the on
the receiving end of the abuse.

01:28:40.400 --> 01:28:42.797
>> Yep. >> Um right.
But [snorts] like software

01:28:42.880 --> 01:28:45.917
software a software engineer is somebody
who basically transmutes his own labor

01:28:46.000 --> 01:28:50.317
into an actual capital asset. Um creates
permanent value. Well, in fact, it's

01:28:50.400 --> 01:28:53.277
actually very inspiring. Um that's
actually more true today than before. So

01:28:53.360 --> 01:28:55.917
when when I was doing software,
the assumption was all new software

01:28:56.000 --> 01:29:00.477
basically has a sort of a parabolic sort
of life cycle, right? So you you ship

01:29:00.560 --> 01:29:03.917
the thing, people buy it. At some point,
everybody who wants it has bought it and

01:29:04.000 --> 01:29:05.357
then it becomes obsolete and it's like

01:29:05.440 --> 01:29:06.957
bananas. Nobody nobody buys old

01:29:07.040 --> 01:29:11.917
software. Um these days um Minecraft um

01:29:12.000 --> 01:29:15.597
Mathematica, you know, Facebook, Google,

01:29:15.680 --> 01:29:19.597
um you have these software assets that
are, you know, have been around for 30

01:29:19.680 --> 01:29:21.677
years that are gaining
in value every year,

01:29:21.760 --> 01:29:23.837
>> right? And they're just they're
being World of Warcraft, right?

01:29:23.920 --> 01:29:26.877
Salesforce.com, like they're being every
single year they're being polished and

01:29:26.960 --> 01:29:27.917
polished and polished and polished.

01:29:28.000 --> 01:29:30.317
They're getting better and better, more
powerful, more powerful, more valuable,

01:29:30.400 --> 01:29:32.877
more valuable. So we we've entered
this era where you can actually have these

01:29:32.960 --> 01:29:35.357
things that actually build out over
decades, which by the way is what's

01:29:35.440 --> 01:29:40.397
happening right now with like GPT.
Um and so um now and and this is why you

01:29:40.480 --> 01:29:43.037
know there there there is always you
know sort of a constant investment

01:29:43.120 --> 01:29:46.317
frenzy around software is because you
know look when you start one of these

01:29:46.400 --> 01:29:49.037
things it doesn't always succeed
but when it does now you might be building

01:29:49.120 --> 01:29:53.517
an asset that builds value for you know
four or five six decades to come. Um you

01:29:53.600 --> 01:29:56.637
know if you have a team of people who
have the level of devotion required to

01:29:56.720 --> 01:30:00.317
keep making it better and then the fact
that of course everybody's online you

01:30:00.400 --> 01:30:03.117
know there's five billion people that
are a click away from any new piece of

01:30:03.200 --> 01:30:06.157
software. So the potential market size
for any of these things is, you know,

01:30:06.240 --> 01:30:09.037
nearly infinite. >> They must have
been surreal back then, though.

01:30:09.120 --> 01:30:11.677
>> Yeah. Yeah. This was all brand new,
right? Yeah. Back then, this was all

01:30:11.760 --> 01:30:14.717
brand new. These were all, you know,
brand new. Had you rolled out that

01:30:14.800 --> 01:30:18.557
theory in even 1999, people would have
thought you were smoking crack. So that

01:30:18.640 --> 01:30:21.357
that's that's emerged over time.

01:30:21.440 --> 01:30:24.717
>> Well, let's uh now turn back into the

01:30:24.800 --> 01:30:27.677
future. You wrote the essay why AI will

01:30:27.760 --> 01:30:31.917
save the world. Let's start at the very
high level. What's the main thesis of

01:30:32.000 --> 01:30:34.957
the essay? Yeah. So, the main thesis
on the essay is that what we're dealing

01:30:35.040 --> 01:30:38.717
with here is intelligence. Um, and it's
really important to kind of talk about

01:30:38.800 --> 01:30:42.157
the sort of very nature of what
intelligence is. And [gasps and sighs]

01:30:42.240 --> 01:30:45.277
fortunately, we have a we have
a predecessor to machine intelligence,

01:30:45.360 --> 01:30:48.717
which is human intelligence. And we've
got, you know, observations and theories

01:30:48.800 --> 01:30:51.997
over thousands of years for what what
what intelligence is in the hands of of

01:30:52.080 --> 01:30:55.197
humans. And and what intelligence is,
right? I mean, what it what it literally

01:30:55.280 --> 01:30:58.237
is is the way to uh, you know,
capture, process, analyze, synthesize

01:30:58.320 --> 01:31:01.037
information, solve problems. Um but the

01:31:01.120 --> 01:31:05.277
observation of of of intelligence
in human hands is that intelligence quite

01:31:05.360 --> 01:31:07.517
literally makes everything better. Um

01:31:07.600 --> 01:31:09.917
and what I mean by that is every kind of

01:31:10.000 --> 01:31:12.557
outcome of like human quality of life

01:31:12.640 --> 01:31:14.557
whether it's education outcomes or

01:31:14.640 --> 01:31:17.197
success of your children or career

01:31:17.280 --> 01:31:20.077
success or health or lifetime

01:31:20.160 --> 01:31:25.037
satisfaction. Um by the way um uh prop
uh uh propensity to peacefulness as

01:31:25.120 --> 01:31:28.317
opposed to violence.
>> Uh propensity for open-mindedness uh

01:31:28.400 --> 01:31:31.437
versus bigotry. Um those are all
associated with higher levels of

01:31:31.520 --> 01:31:33.597
intelligence. >> Smarter
people have better outcomes in

01:31:33.680 --> 01:31:36.797
almost as you write in almost
every domain of activity. Academic

01:31:36.880 --> 01:31:40.397
achievement, job performance,
occupational status, income, creativity,

01:31:40.480 --> 01:31:42.557
physical health, longevity, learning new

01:31:42.640 --> 01:31:47.437
skills, managing complex tasks,
leadership, entrepreneurial success,

01:31:47.520 --> 01:31:52.077
conflict resolution, reading
comprehension, financial decision-m,

01:31:52.160 --> 01:31:55.037
understanding others perspectives,
creative arts, parenting outcomes, and

01:31:55.120 --> 01:32:00.557
life satisfaction. one of the more
depressing conversations I've had and I

01:32:00.640 --> 01:32:02.797
don't know why it's depressing.
I have to really think through why it's

01:32:02.880 --> 01:32:06.877
depressing but on IQ

01:32:06.960 --> 01:32:10.237
and uh the G factor

01:32:10.320 --> 01:32:15.117
and that that's something
in large part is genetic.

01:32:15.200 --> 01:32:16.477
>> Mhm.

01:32:16.560 --> 01:32:19.437
>> And it correlates so much with all of

01:32:19.520 --> 01:32:22.397
these things and success in life. It's

01:32:22.480 --> 01:32:25.197
like all the inspirational stuff we read

01:32:25.280 --> 01:32:28.477
about like if you work hard and so on.

01:32:28.560 --> 01:32:32.317
Damn, it sucks that you're born
with a hand that you can't change.

01:32:32.400 --> 01:32:36.157
>> But what if you could? you're you're
saying basically a really important

01:32:36.240 --> 01:32:38.957
point and I think it's a uh in in your

01:32:39.040 --> 01:32:40.717
articles

01:32:40.800 --> 01:32:44.477
it it really helped me um it's a nice

01:32:44.560 --> 01:32:49.037
added perspective to think about
listen human intelligence the science of

01:32:49.120 --> 01:32:51.197
intelligence has shown scientifically

01:32:51.280 --> 01:32:53.517
that it just makes life easier and

01:32:53.600 --> 01:32:57.677
better the smarter you are and now let's

01:32:57.760 --> 01:33:00.397
look at artificial intelligence

01:33:00.480 --> 01:33:06.077
and if uh that's a way to increase the

01:33:06.160 --> 01:33:10.637
the the some human intelligence then
it's only going to make a better life.

01:33:10.720 --> 01:33:12.397
Yeah, that's the argument.
And [clears throat] certainly at the

01:33:12.480 --> 01:33:14.637
collective level we could talk about
the collective effect of just having more

01:33:14.720 --> 01:33:18.397
intelligence in the world which which
will have very big payoff but there's

01:33:18.480 --> 01:33:21.917
also just at the individual level like
what if every person has a machine you

01:33:22.000 --> 01:33:26.397
know and the concept of augment Doug
Engelar's concept of augmentation um you

01:33:26.480 --> 01:33:28.477
know what if everybody has a an

01:33:28.560 --> 01:33:31.117
assistant and the assistant is you know

01:33:31.200 --> 01:33:35.917
140 IQ um and you happen to be 110 IQ um

01:33:36.000 --> 01:33:39.437
and you've got you know something
that basically is infinitely patient and

01:33:39.520 --> 01:33:43.357
knows everything about you and is
pulling for you in every possible way,

01:33:43.440 --> 01:33:47.277
wants you to be successful and anytime
you find anything confusing or want to

01:33:47.360 --> 01:33:49.357
learn anything or have trouble
understanding something or want to

01:33:49.440 --> 01:33:52.317
figure out what to do in a situation,
right? Want to figure out how to prepare

01:33:52.400 --> 01:33:55.917
for a job interview like any of these
things like it will help you do it and

01:33:56.000 --> 01:34:00.397
it will therefore the combination will
effectively be you know effectively

01:34:00.480 --> 01:34:03.517
raise your raise because it will
effectively raise your IQ will therefore

01:34:03.600 --> 01:34:06.397
raise the odds of of successful life
outcomes in all these areas. So people

01:34:06.480 --> 01:34:09.997
below the this hypothetical 140 IQ,

01:34:10.080 --> 01:34:12.237
it'll pull them up towards 140 IQ.

01:34:12.320 --> 01:34:15.517
>> Yeah. Yeah. Yeah. And then of course,
you know, people at people at 140 IQ

01:34:15.600 --> 01:34:18.557
will be able to have a peer, right,
to be able to commun, which is great. And

01:34:18.640 --> 01:34:21.437
then people above 140 IQ will have
an assistant that they can farm things out

01:34:21.520 --> 01:34:24.877
to. And then look, God willing, you know,
at some point these things go from

01:34:24.960 --> 01:34:28.957
future versions go from 140 IQ
equivalent to 150 to 160 to [laughter]

01:34:29.040 --> 01:34:31.357
180, right? Like Einstein was estimated

01:34:31.440 --> 01:34:33.757
to be on the order of 160. um you know

01:34:33.840 --> 01:34:39.117
so when we get you know 160 AI like
we'll be you know when one assumes

01:34:39.200 --> 01:34:43.117
creating Einstein level breakthroughs
and physics and and then and then at 180

01:34:43.200 --> 01:34:46.877
we'll be you know curing cancer
and developing warp drive and doing all

01:34:46.960 --> 01:34:50.397
kinds of stuff and so it is quite
possibly the case this is the most

01:34:50.480 --> 01:34:52.587
important thing that's ever happened
and the best thing that's ever happened

01:34:52.670 --> 01:34:56.237
[snorts] because precisely because
it's a lever on this single fundamental

01:34:56.320 --> 01:34:58.957
factor of intelligence which is the thing
that drives so much of everything

01:34:59.040 --> 01:35:04.397
else. Can you steal man the case that
human plus AI is not always better than

01:35:04.480 --> 01:35:06.797
human for the individual?
>> You may have noticed that there's a lot

01:35:06.880 --> 01:35:09.277
of smart running around. >> Sure. Yes.

01:35:09.360 --> 01:35:12.877
>> Right. And so like smart there are
certain people where they get smarter,

01:35:12.960 --> 01:35:16.237
you know, they get to be more arrogant,
right? So, you know, there's one huge

01:35:16.320 --> 01:35:19.517
flaw. >> Although to push
back on that, it might

01:35:19.600 --> 01:35:23.517
be interesting because when the
intelligence is not all coming from you,

01:35:23.600 --> 01:35:27.677
but from a from another system, that
might actually increase the the amount

01:35:27.760 --> 01:35:31.597
of humility even in the
One would hope. Yeah. Um

01:35:31.680 --> 01:35:34.797
>> or it could make more You know,
that's I mean that's that's

01:35:34.880 --> 01:35:38.797
for psychology to study. >> Yeah.
Exactly. Another one is um smart

01:35:38.880 --> 01:35:41.597
people are very convinced that they,
you know, have a more rational view of the

01:35:41.680 --> 01:35:44.477
world and that they have a easier time
seeing through conspiracy theories and

01:35:44.560 --> 01:35:47.597
hoaxes and right, you know,
sort of crazy beliefs and all that. There

01:35:47.680 --> 01:35:51.037
there's a theory in psychology which
is actually smart people. So for sure

01:35:51.120 --> 01:35:53.677
people who aren't as smart are very
susceptible to hoaxes and conspiracy

01:35:53.760 --> 01:35:56.957
theories. But it may also be the case
that the smarter you get, you become

01:35:57.040 --> 01:36:00.957
susceptible in a different way. uh which
is you become very good at marshalling

01:36:01.040 --> 01:36:03.357
facts to fit preconceptions. >> Yes.

01:36:03.440 --> 01:36:07.277
>> Right. Um you become very very good
at assembling whatever theories and

01:36:07.360 --> 01:36:11.277
frameworks and pieces of data and graphs
and charts you need to validate whatever

01:36:11.360 --> 01:36:12.957
crazy ideas got into your head.

01:36:13.040 --> 01:36:16.397
>> And so you're susceptible
in a different way, right?

01:36:16.480 --> 01:36:20.157
>> Uh we're all sheep but [laughter]
different colored sheep.

01:36:20.240 --> 01:36:23.677
>> Some sheep are better at justifying it,
right? Um and those are, you know, those

01:36:23.760 --> 01:36:27.677
are the smart sheep, right? Um so yeah
look like I would say this look like

01:36:27.760 --> 01:36:31.037
there are no panace I am not I am not
a utopian. There are no panaceas in life.

01:36:31.120 --> 01:36:34.157
Um there are no like you know I don't
believe there are like pure positives.

01:36:34.240 --> 01:36:37.997
I'm not a transcendental kind of person
like that. But you know so yeah there

01:36:38.080 --> 01:36:41.597
are going to be issues. Um uh and um
and you know look smart people. Another

01:36:41.680 --> 01:36:43.837
thing maybe you could say about smart
people is they are more likely to get

01:36:43.920 --> 01:36:46.317
themselves in situations that are you
know beyond their grasp you know because

01:36:46.400 --> 01:36:49.197
they're just more confident in their
ability to deal with complexity and

01:36:49.280 --> 01:36:52.077
their their eyes become bigger. Their
their cognitive eyes become bigger than

01:36:52.160 --> 01:36:54.637
their stomach >> you know.
So yeah, you could argue those

01:36:54.720 --> 01:36:59.277
eight different ways. Nevertheless,
on net, right, clearly overwhelmingly

01:36:59.360 --> 01:37:02.077
again, if you just extrapolate from what
we know about human intelligence, you're

01:37:02.160 --> 01:37:04.397
you're improving so many aspects of life

01:37:04.480 --> 01:37:06.557
if you're upgrading intelligence.

01:37:06.640 --> 01:37:11.597
>> So there'll be assistance at all stages
of life. So when you're younger, there's

01:37:11.680 --> 01:37:16.477
for education, all that kind of stuff,
and mentorship, all all of this. And uh

01:37:16.560 --> 01:37:19.677
later on as you're doing work and you've
developed a skill and you're having a

01:37:19.760 --> 01:37:23.277
profession, you'll have an assistant
that helps you excel at that profession.

01:37:23.360 --> 01:37:25.837
So at all stages of life.
>> Yeah. I mean look the theory is

01:37:25.920 --> 01:37:29.277
augmentations. This is the Doug
Eglebert's terart made this observation

01:37:29.360 --> 01:37:31.997
many many decades ago that you know
basically it's like you can have this

01:37:32.080 --> 01:37:34.797
oppositional frame of technology where
it's like us versus the machines. But

01:37:34.880 --> 01:37:37.757
what you really do is you use
technology to augment human capabilities.

01:37:37.840 --> 01:37:39.117
>> Yeah. >> And and then
by the way that's how

01:37:39.200 --> 01:37:41.677
actually the economy develops. That's
we can talk about the economic side of

01:37:41.760 --> 01:37:44.477
this, but that that's actually how
the economy grows um is through through

01:37:44.560 --> 01:37:47.277
technology augmenting
human human potential.

01:37:47.360 --> 01:37:49.517
>> Um and so yeah, and then you you

01:37:49.600 --> 01:37:54.317
basically have a a proxy or you know
or or a um you know, a sort of prosthetic.

01:37:54.400 --> 01:37:57.597
Um you know, so like you've got glasses,
you've got a wristwatch, you know,

01:37:57.680 --> 01:38:00.637
you've got shoes, you you know you've
got these things, you've got a personal

01:38:00.720 --> 01:38:04.237
computer, you've got a word processor,
you've got Mathematica, you've got

01:38:04.320 --> 01:38:08.797
Google. This is the late viewed through
that lens. AI is the latest in a long

01:38:08.880 --> 01:38:12.157
series of basically augmentation
methods uh to be able to raise human

01:38:12.240 --> 01:38:15.357
capabilities. It's just this one is the
most powerful one of all because this is

01:38:15.440 --> 01:38:18.477
the one that that goes directly to what
what they call fluid intelligence

01:38:18.560 --> 01:38:21.437
>> which is IQ.

01:38:21.520 --> 01:38:24.637
>> Well, there's uh
two categories of folks

01:38:24.720 --> 01:38:29.517
that you outline that uh that worry
about or highlight the risks of AI and

01:38:29.600 --> 01:38:33.117
you highlight a bunch of different risks.
I would love to go through those

01:38:33.200 --> 01:38:35.997
risks and just discuss them, brainstorm

01:38:36.080 --> 01:38:40.237
which ones are serious and which ones
are less serious. But first, the the

01:38:40.320 --> 01:38:41.757
Baptist and the bootleggers. What are

01:38:41.840 --> 01:38:44.397
these two interesting groups of folks

01:38:44.480 --> 01:38:49.037
who uh who who worry about uh the effect

01:38:49.120 --> 01:38:51.757
of AI on human civilization
>> or say they do?

01:38:51.840 --> 01:38:55.037
>> Say okay. [laughter] Yes.
Or say they do.

01:38:55.120 --> 01:38:57.357
>> The Baptists worry,
the bootleggers say they do.

01:38:57.440 --> 01:39:01.117
>> Yeah. Um so the Baptist and the
bootleggers is a metaphor from economics

01:39:01.200 --> 01:39:03.837
um from what's called development
economics and it's this observation that

01:39:03.920 --> 01:39:08.797
when you get social reform movements um
in a society um you tend to get two sets

01:39:08.880 --> 01:39:12.717
of people showing up arguing
for the social reform um and the the term

01:39:12.800 --> 01:39:15.677
Baptist and bootleggers comes from
the American experience with alcohol

01:39:15.760 --> 01:39:21.037
prohibition um and so in the 1900s 1910s
um there was this movement that was very

01:39:21.120 --> 01:39:24.637
passionate at the time which basically
said alcohol is evil uh and it's

01:39:24.720 --> 01:39:28.637
destroying society. Um by the way there
was a lot of evidence to support this.

01:39:28.720 --> 01:39:33.277
Um there was very high rates of uh very
high correlations then by the way and

01:39:33.360 --> 01:39:37.357
now uh between rates of physical
violence and alcohol use. Um almost all

01:39:37.440 --> 01:39:40.157
violent crimes have either
the perpetrator or the victim are both

01:39:40.240 --> 01:39:43.917
drunk. Almost if you see this actually
in the work almost all sexual harassment

01:39:44.000 --> 01:39:47.197
cases in the workplace. It's like at a
company party and somebody's drunk. Like

01:39:47.280 --> 01:39:50.157
it's it's amazing how often alcohol
actually correlates to actually

01:39:50.240 --> 01:39:54.237
dysfunction. It leads to domestic abuse
um and so forth, child abuse. And so you

01:39:54.320 --> 01:39:56.797
had this group of people who were like,
"Okay, this this is bad stuff and we

01:39:56.880 --> 01:39:59.757
should outlaw it." And and those were
quite literally Baptists. Those were

01:39:59.840 --> 01:40:03.117
super committed, you know, hardcore
Christian activists in a lot of cases.

01:40:03.200 --> 01:40:07.117
There was this woman uh whose name
was Carrie Nation, um who was this older

01:40:07.200 --> 01:40:09.597
woman who had been in this, you know,
I don't know, disastrous marriage or

01:40:09.680 --> 01:40:12.637
something and her husband had been
abusive and drunk all the time. She

01:40:12.720 --> 01:40:14.957
became the icon of the Baptist uh

01:40:15.040 --> 01:40:19.917
prohibitionist and she was legendary
in that era for carrying an axe um and

01:40:20.000 --> 01:40:23.357
doing you know completely on her own
doing raids of saloons and like taking

01:40:23.440 --> 01:40:27.997
her axe to all the bottles and tags >> in
the back and and so >> so a true believer

01:40:28.080 --> 01:40:31.997
>> an absolute true believer um and with
absolutely the purest of intentions and

01:40:32.080 --> 01:40:35.917
and again there's a very important thing
here which is there's you could look at

01:40:36.000 --> 01:40:38.397
this cynically and you could say
the Baptists are like delusional you know

01:40:38.480 --> 01:40:41.517
extremists but you could also say look
they're right like she was you know she

01:40:41.600 --> 01:40:45.437
had a point [snorts] >> like she wasn't
wrong about a lot of what she said.

01:40:45.520 --> 01:40:47.677
>> Yeah. >> But it turns
out the way the story goes

01:40:47.760 --> 01:40:50.797
is it turns out that there were another
set of people who very badly wanted to

01:40:50.880 --> 01:40:54.157
outlaw alcohol in those days and those
were the bootleggers which was organized

01:40:54.240 --> 01:40:58.557
crime that stood to make a huge amount
of money if legal alcohol sales were

01:40:58.640 --> 01:41:01.677
banned. Um and this was in fact the way
the history goes is this was actually

01:41:01.760 --> 01:41:04.157
the beginning of organized crime
in the US. This was the big economic

01:41:04.240 --> 01:41:08.957
opportunity that opened that up. Um and so
they went in together. Um, and I know

01:41:09.040 --> 01:41:11.997
they didn't go in together like
the Baptists did not even necessarily know

01:41:12.080 --> 01:41:14.317
about the bootleggers because
they were on their moral crusade.

01:41:14.400 --> 01:41:15.597
>> The bootleggers
certainly knew about the

01:41:15.680 --> 01:41:18.237
Baptists and they were like, "Wow, this
is these people are like the great front

01:41:18.320 --> 01:41:21.677
people for like, you know, >> good
>> shenanigans in the background."

01:41:21.760 --> 01:41:24.157
>> Yeah. >> And they got
the Folstead Act passed,

01:41:24.240 --> 01:41:27.837
right? And they did in fact ban alcohol
in the US. And you'll notice what

01:41:27.920 --> 01:41:30.077
happened, which is people kept drinking.

01:41:30.160 --> 01:41:33.837
>> It didn't work. [laughter] People
kept drinking. Um, that bootleggers made a

01:41:33.920 --> 01:41:37.117
tremendous amount of money. Um, and then
over time it became clear that it made

01:41:37.200 --> 01:41:40.157
no sense to make it illegal and it
was causing more problems. And so then it

01:41:40.240 --> 01:41:43.197
was revoked and here we sit with legal
alcohol 100 years later with all the

01:41:43.280 --> 01:41:46.477
same problems. >> Mh. >> Um,
and you know the whole thing was

01:41:46.560 --> 01:41:50.237
this like giant misadventure. U
the Baptists got taken advantage of by the

01:41:50.320 --> 01:41:53.037
bootleggers and the bootleggers got
what they wanted and and that was it.

01:41:53.120 --> 01:41:57.757
>> The same two categories of folks are
now uh sort of suggesting that uh the

01:41:57.840 --> 01:41:59.997
development of artificial
intelligence should be regulated

01:42:00.080 --> 01:42:02.477
>> 100%. Yeah, it's the same pattern
and the the economists would tell you it's

01:42:02.560 --> 01:42:04.877
the same pattern every time. like this
is what happened at nuclear power. This

01:42:04.960 --> 01:42:07.677
is what happened which is another
interesting one but like yeah this is

01:42:07.760 --> 01:42:11.357
this happens dozens and dozens of times
um throughout the last hundred years and

01:42:11.440 --> 01:42:12.957
and and this is what's happening now.

01:42:13.040 --> 01:42:18.157
>> And you write that it isn't sufficient
to simply identify the actors and impugn

01:42:18.240 --> 01:42:21.357
their motives. We should consider
the arguments of both the Baptists and the

01:42:21.440 --> 01:42:23.117
bootleggers on their merits.

01:42:23.200 --> 01:42:29.320
>> So let's do just that.
>> Risk number one

01:42:29.520 --> 01:42:35.027
uh will AI kill us all? >> Yes. So

01:42:35.110 --> 01:42:37.037
[clears throat] uh what do
you what do you think about

01:42:37.120 --> 01:42:39.117
this one? This this what do you think is

01:42:39.200 --> 01:42:42.397
the core argument here

01:42:42.480 --> 01:42:47.917
that uh the development of AGI perhaps
better said uh will destroy human

01:42:48.000 --> 01:42:49.997
civilization? >> Well,
first of all, you just did a

01:42:50.080 --> 01:42:54.017
slight of hand because we went
from talking about AI to AGI.

01:42:54.100 --> 01:42:55.997
[laughter] >> Is there
a fundamental difference there?

01:42:56.080 --> 01:42:59.517
>> I don't know. What's AGI?
>> I What's AI? What's in

01:42:59.600 --> 01:43:02.877
>> I know what AI is.
AI is machine learning. What's what's AGI?

01:43:02.960 --> 01:43:06.077
>> I think we don't know what the bottom
of the well of machine learning is or what

01:43:06.160 --> 01:43:10.317
the ceiling is because just uh to call
something machine learning or just to

01:43:10.400 --> 01:43:14.477
call something statistics or just to call
it math or computation doesn't mean

01:43:14.560 --> 01:43:16.477
you know uh nuclear weapons or just

01:43:16.560 --> 01:43:21.997
physics. So it's it it's to me it's
very interesting and surprising how far

01:43:22.080 --> 01:43:23.997
machine learning has taken. >> No,
but we knew that nuclear physics

01:43:24.080 --> 01:43:27.037
would lead to weapons. That's why
the scientists of that era were always in

01:43:27.120 --> 01:43:29.597
this huge dispute about building
the weapons. This is different. Asia is

01:43:29.680 --> 01:43:31.597
different. >> Where does machine
learning lead? Do we know?

01:43:31.680 --> 01:43:34.317
>> We don't know. But this my point
is different. We we actually don't know.

01:43:34.400 --> 01:43:37.357
But and and this is where you the slight
of hand kicks in, right? This is where

01:43:37.440 --> 01:43:41.437
it goes from being a scientific topic
to being a religious topic. Um and that's

01:43:41.520 --> 01:43:43.917
that's why I specifically called out
that because that's what happens. They

01:43:44.000 --> 01:43:46.077
do the vocabulary shift and all of a
sudden you're talking about something

01:43:46.160 --> 01:43:47.597
totally that's not actually real.

01:43:47.680 --> 01:43:50.237
>> Well, then maybe
you can also uh as part

01:43:50.320 --> 01:43:54.077
of that define the western
tradition of millennism.

01:43:54.160 --> 01:43:56.477
>> Yes. End of the world. Apocalypse.

01:43:56.560 --> 01:44:01.037
Apocalypse. Apocalypse cults. Um >>
apocalypse cults. Well, so we live in we

01:44:01.120 --> 01:44:03.917
of course live in a Judeo-Christian
but primarily Christian kind of saturated

01:44:04.000 --> 01:44:06.397
you know kind of Christian
post-Christian secularized Christian you

01:44:06.480 --> 01:44:10.477
know kind of world in the west. Um
and of course core to Christianity is the

01:44:10.560 --> 01:44:14.237
idea of the second coming and and you
know revelations and you know the Jesus

01:44:14.320 --> 01:44:17.437
returning and thou the thousand-year
you know utopia on earth and then the you

01:44:17.520 --> 01:44:20.557
know the rapture and like all all that
stuff you know we don't we you know we

01:44:20.640 --> 01:44:23.357
collectively you know as a society we
don't necessarily take all that fully

01:44:23.440 --> 01:44:27.117
seriously now. So what we do is we
create our secularized versions of that.

01:44:27.200 --> 01:44:30.237
We keep we keep looking for utopia.
we keep looking for, you know, basically

01:44:30.320 --> 01:44:33.517
the end of the world. And and so what
what you see over over decades is a

01:44:33.600 --> 01:44:37.197
basically a pattern of these sort of of
these of these of is this this is what

01:44:37.280 --> 01:44:39.997
cults are. This is how cults form
is they form around some theory of the end

01:44:40.080 --> 01:44:41.597
of the world. And so the People's Temple

01:44:41.680 --> 01:44:47.117
cult, the Manson cult, the Heavensgate
cult, the David Caresh cult. You know

01:44:47.200 --> 01:44:49.437
what they're all organized around
is like there's going to be this thing

01:44:49.520 --> 01:44:52.157
that's going to happen that's going
to basically bring civilization crashing

01:44:52.240 --> 01:44:55.117
down. And then we have this special
elite group of people who are going to

01:44:55.200 --> 01:44:58.157
see it coming and prepare for it.
And then they're the people who are either

01:44:58.240 --> 01:45:00.957
going to stop it or are failing stopping
it. They're going to be the people who

01:45:01.040 --> 01:45:03.757
survive to the other side and ultimately
get credit for having been right.

01:45:03.840 --> 01:45:05.437
>> Why is that so
compelling, do you think?

01:45:05.520 --> 01:45:07.997
Like um because it satisfies this very

01:45:08.080 --> 01:45:13.757
deep need we have for transcendence
and meaning that got stripped away when we

01:45:13.840 --> 01:45:16.637
became secular. >> Yeah.
But why why does the transcendence

01:45:16.720 --> 01:45:18.477
involve the destruction of human

01:45:18.560 --> 01:45:23.997
civilization? Cuz like how like how
plausible Well, it's it's like a very

01:45:24.080 --> 01:45:26.557
deep psychological thing because it's
like how plausible how plausible is it

01:45:26.640 --> 01:45:29.357
that we live in a world where
everything's just kind of all right,

01:45:29.440 --> 01:45:32.157
>> right? How exciting >>
how exciting is that? Right.

01:45:32.240 --> 01:45:36.557
>> But that's more than that. >>
But that's the deep question I'm asking.

01:45:36.640 --> 01:45:40.557
>> Why is it not exciting to live in a
world where everything's just all right?

01:45:40.640 --> 01:45:44.477
Cuz >> I think u you
know most of the animal

01:45:44.560 --> 01:45:46.397
kingdom would be so happy with just all

01:45:46.480 --> 01:45:49.117
right cuz that means survival. Why are

01:45:49.200 --> 01:45:51.837
we uh maybe that's what it is. Why are

01:45:51.920 --> 01:45:55.037
we conjuring up things to worry about?

01:45:55.120 --> 01:45:59.437
>> So CS Lewis called it the God-shaped
hole. So there's a God-shaped hole in

01:45:59.520 --> 01:46:03.197
the human experience, consciousness,
soul, whatever you want to call it,

01:46:03.280 --> 01:46:06.957
where there's got to be something
that's bigger than all this.

01:46:07.040 --> 01:46:08.637
>> There's got to be something
transcendent. There's got to be

01:46:08.720 --> 01:46:12.477
something that is bigger, right? Bigger,
a bigger purpose, a bigger meaning. And

01:46:12.560 --> 01:46:15.917
so we have run the experiment of,
you know, we're just going to use science

01:46:16.000 --> 01:46:18.797
and rationality and kind of, you know,
everything's just going to kind of be as

01:46:18.880 --> 01:46:21.357
it appears. And large number of people

01:46:21.440 --> 01:46:25.997
have found that very deeply wanting
and have constructed narratives and and by

01:46:26.080 --> 01:46:28.717
this is the story of the 20th century,
right? Communism, right, was one of

01:46:28.800 --> 01:46:33.357
those. Communism was a was a form of this.
Nazism was a form of this. Um, you

01:46:33.440 --> 01:46:36.797
know, some people um you know, you can
see movements like this playing out all

01:46:36.880 --> 01:46:40.237
over the world right now. >> So you
construct a kind of devil, a kind

01:46:40.320 --> 01:46:43.437
of source of evil. and we're
going to transcend beyond it.

01:46:43.520 --> 01:46:47.757
>> Yeah. And the millinarian,
the millinarians kind of when you see a

01:46:47.840 --> 01:46:50.797
millinarian cult, they put a really
specific point on it, which is end of

01:46:50.880 --> 01:46:54.717
the world, right? There there is some
change coming and that change that's

01:46:54.800 --> 01:46:58.317
coming is so profound and so important
that it's either going to lead to utopia

01:46:58.400 --> 01:47:03.117
or hell on earth, >> right?
Um and it is going to and then

01:47:03.200 --> 01:47:06.157
you know it's like what if you actually
knew that that was going to happen,

01:47:06.240 --> 01:47:09.437
right? What would you what what would
you do, right? How would you prepare

01:47:09.520 --> 01:47:12.157
yourself for it? How would you come
together with a group of like-minded

01:47:12.240 --> 01:47:15.517
people, right? How would you what would
you do? Would you plan like hashes of

01:47:15.600 --> 01:47:18.637
weapons in the woods? Would you like,
you know, I don't know, create under

01:47:18.720 --> 01:47:21.277
underground buckers? Would you,
you know, spend your life trying to figure

01:47:21.360 --> 01:47:22.877
out a way to avoid having it happen?

01:47:22.960 --> 01:47:25.837
>> Yeah, that's a really
compelling, exciting

01:47:25.920 --> 01:47:30.797
idea to uh to have a club over to have
to to have a to have a little bit of

01:47:30.880 --> 01:47:33.917
tribe like a get together on a Saturday
night and drink some beers and talk

01:47:34.000 --> 01:47:36.797
about the the end of the world and how
you're the you [clears throat] are the

01:47:36.880 --> 01:47:38.397
only ones who have figured it out.

01:47:38.480 --> 01:47:41.197
>> Yeah. And then and then once you lock
in on that, like how can you do anything

01:47:41.280 --> 01:47:43.437
else with your life? Like this is
obviously the thing that you have to do.

01:47:43.520 --> 01:47:45.757
And then and then there's
a psychological effect that you alluded

01:47:45.840 --> 01:47:48.317
to. There's a psychological effect. If
you take a set of true believers and you

01:47:48.400 --> 01:47:50.797
leave them to themselves, they get
more radical, right? Because they they

01:47:50.880 --> 01:47:53.917
self-radicalize each other.
>> That said, >> yes,

01:47:54.000 --> 01:47:56.477
>> it doesn't mean they're
not sometimes right.

01:47:56.560 --> 01:47:57.917
>> Yeah.
The end of the world might be Yes.

01:47:58.000 --> 01:47:59.437
Correct. Like they might be right.

01:47:59.520 --> 01:48:01.867
>> Yeah. >> But [snorts] like we
>> I have some pamphlets for you.

01:48:01.950 --> 01:48:05.197
[laughter] >> Exactly.
It's it I mean there's I mean

01:48:05.280 --> 01:48:08.157
we'll talk about nuclear weapons because
you have a really interesting little

01:48:08.240 --> 01:48:12.077
moment that I learned about in in your
essay but you know sometimes it could be

01:48:12.160 --> 01:48:15.117
right. >> Yeah. >> Because
we're still you we're developing

01:48:15.200 --> 01:48:19.437
more and more powerful technologies uh
in this case and we don't know what the

01:48:19.520 --> 01:48:22.877
impact it will have on human civilization.
Well we can highlight all

01:48:22.960 --> 01:48:27.117
the different predictions about how it
will be positive but the risks are there

01:48:27.200 --> 01:48:30.717
and you discuss some of them. Well,
the steel man, the steel man is the steel

01:48:30.800 --> 01:48:33.677
man actually the steel man and his
reputation are the same which is you

01:48:33.760 --> 01:48:37.197
can't predict what's going to happen
right you right you can't rule out that

01:48:37.280 --> 01:48:40.877
this will not end everything right
but the response to that is you have just

01:48:40.960 --> 01:48:44.077
made a completely non-scientific claim
>> you've made a religious claim not a

01:48:44.160 --> 01:48:45.917
scientific claim there >>
how does it get disproven

01:48:46.000 --> 01:48:48.477
>> there is and there's no by definition
with these kinds of claims there's no

01:48:48.560 --> 01:48:51.997
way to disprove them >> right
um and so there there's no you go

01:48:52.080 --> 01:48:55.917
right on the list there's no hypothesis
there's no testability of the hypothesis

01:48:56.000 --> 01:48:58.957
there's no um way
to falsify the hypothesis

01:48:59.040 --> 01:49:01.117
there's no way to measure progress along

01:49:01.200 --> 01:49:06.237
the arc. Like it's just all completely
missing and so it's not scientific and

01:49:06.320 --> 01:49:09.437
>> well I don't I don't think it's
completely missing. It's it's somewhat

01:49:09.520 --> 01:49:12.877
missing. So for example the the the the
people that say AI is going to kill all

01:49:12.960 --> 01:49:15.917
of us. I mean they usually have ideas

01:49:16.000 --> 01:49:20.717
about how to do that whether it's
the paperclip maximizer or um you know it

01:49:20.800 --> 01:49:25.437
escapes. There's mechanism by which
you can imagine it killing all humans

01:49:25.520 --> 01:49:28.237
>> models >> and

01:49:28.320 --> 01:49:30.877
to you can disprove it by saying there

01:49:30.960 --> 01:49:35.357
is um there is a limit to uh the speed

01:49:35.440 --> 01:49:38.477
at which intelligence increases

01:49:38.560 --> 01:49:42.637
maybe show that uh like sort of

01:49:42.720 --> 01:49:45.997
rigorously really describe model like

01:49:46.080 --> 01:49:51.037
how it could happen and say no there
here's a physics limitation there's a

01:49:51.120 --> 01:49:54.637
physical limitation to how these
systems would actually do damage to human

01:49:54.720 --> 01:49:56.717
civilization. And it is possible they

01:49:56.800 --> 01:50:02.237
will kill 10 to 20% of the population,
but it seems impossible for them to kill

01:50:02.320 --> 01:50:04.557
uh 99%. >> There's
practical counter arguments,

01:50:04.640 --> 01:50:07.277
right? So you mentioned basically what
I described as the thermodynamic counter

01:50:07.360 --> 01:50:10.557
argument, which sitting here today,
it's like where would the evil AGI get the

01:50:10.640 --> 01:50:12.877
GPUs? >> Yeah.
>> Cuz like they don't exist.

01:50:12.960 --> 01:50:15.597
>> So you're going to have a very
frustrated baby evil AGI who's going to

01:50:15.680 --> 01:50:17.837
be like trying to buy
Nvidia stock or something to

01:50:17.920 --> 01:50:21.277
>> get them to finally make some chips.
Um right. So the the serious form of that

01:50:21.360 --> 01:50:23.757
is the thermodynamic argument which
is like okay where's the energy going to

01:50:23.840 --> 01:50:26.557
come from where's the processor going
to be running where is the data center

01:50:26.640 --> 01:50:28.877
going to be happening how is this going
to be happening in secret such that you

01:50:28.960 --> 01:50:32.477
know it's not you know so so that's
a practical counterargument to the runaway

01:50:32.560 --> 01:50:35.917
AGI thing I have a but I have a and we
can argue that and discuss that I have I

01:50:36.000 --> 01:50:39.437
have a deeper objection to it which
is it's this is all forecasting it's all

01:50:39.520 --> 01:50:41.357
modeling it's all it's all future

01:50:41.440 --> 01:50:46.477
prediction it's all future
hypothesizing >> it's not science

01:50:46.560 --> 01:50:49.117
>> sure >> it is not it
is it is is the opposite of

01:50:49.200 --> 01:50:51.997
science So the I'll pull up Carl
Sean extraordinary claims require

01:50:52.080 --> 01:50:56.237
extraordinary proof right these are
extraordinary claims the policies that

01:50:56.320 --> 01:51:00.637
are being called for right to prevent
this are of extraordinary magnitude and

01:51:00.720 --> 01:51:03.437
I think we're going to cause
extraordinary damage and this is all

01:51:03.520 --> 01:51:06.317
being done on the basis of something
that is literally not scientific it's

01:51:06.400 --> 01:51:08.957
not a testable hypothesis >> so
the moment you say AI is going to

01:51:09.040 --> 01:51:12.717
kill all of us therefore we should ban
it or that we should uh regulate all

01:51:12.800 --> 01:51:14.557
that kind of stuff that's
when it starts getting serious

01:51:14.640 --> 01:51:17.117
>> or start you know military
air strikes and data centers

01:51:17.200 --> 01:51:21.437
>> oh boy Right. And like, [laughter]

01:51:21.520 --> 01:51:24.317
>> yeah, that's when it get
starts it starts getting real.

01:51:24.400 --> 01:51:25.437
>> So, here's the problem
with millionarian

01:51:25.520 --> 01:51:28.957
cults. They have a hard time
staying away from violence.

01:51:29.040 --> 01:51:30.817
>> Yeah. But violence is so fun, man.

01:51:30.900 --> 01:51:35.277
[laughter] >> Well, if you're
on the right end of it, they

01:51:35.360 --> 01:51:37.357
have a hard time avoiding violence.
The reason they have a hard time avoiding

01:51:37.440 --> 01:51:40.637
violence is if you
actually believe the claim,

01:51:40.720 --> 01:51:44.317
>> right? Then what would you do to stop
the end of the world? Well, you would do

01:51:44.400 --> 01:51:47.997
anything, right? And so and this is where
you get and again if you just look

01:51:48.080 --> 01:51:50.717
at the history of of millinarian cults
this is where you get the people's

01:51:50.800 --> 01:51:52.957
temple and everybody killing themselves
in the jungle and this is where you get

01:51:53.040 --> 01:51:57.597
Charles Manson and you know sending
in to kill kill the pigs like this is the

01:51:57.680 --> 01:52:00.797
problem with these they they have a very
hard time to run the line at actual

01:52:00.880 --> 01:52:04.477
violence and I think I think in this
case there's they're I mean they're

01:52:04.560 --> 01:52:08.157
already calling for it like today
and you know where this goes from here as

01:52:08.240 --> 01:52:11.757
they get more worked up like I I think
is like really concerning. Okay, but

01:52:11.840 --> 01:52:15.597
that's kind of the extremes. You know,
the extremes of anything are always

01:52:15.680 --> 01:52:18.877
concerning. It it's also
possible to kind of believe

01:52:18.960 --> 01:52:20.957
that AI has a very high likelihood of

01:52:21.040 --> 01:52:24.637
killing all of us. Uh but there's and

01:52:24.720 --> 01:52:27.837
therefore we should uh maybe consider

01:52:27.920 --> 01:52:32.317
uh slowing development or regulating.
So not violence or any of these kinds of

01:52:32.400 --> 01:52:36.317
things, but saying like all right, let's
let's take a pause here. You know, you

01:52:36.400 --> 01:52:39.757
biological weapons, nuclear weapons,
like whoa, whoa, whoa, whoa, whoa, whoa.

01:52:39.840 --> 01:52:42.797
This is like serious stuff. We should be

01:52:42.880 --> 01:52:47.917
careful. So it is possible to kind of have
a more rational response, right? If

01:52:48.000 --> 01:52:50.877
you believe this risk is real, >> believe.

01:52:50.960 --> 01:52:55.677
>> Yes. So what is it possible to be
have a scientific approach to the the the

01:52:55.760 --> 01:52:58.637
prediction of the future? >> I mean,
we just went through this with CO. Yeah.

01:52:58.720 --> 01:53:01.037
>> What do we know about modeling?

01:53:01.120 --> 01:53:04.077
>> Well, I mean, >> what do we
learn about modeling with CO?

01:53:04.160 --> 01:53:05.677
>> Uh there's a lot of lessons.

01:53:05.760 --> 01:53:09.037
>> They didn't work at all.
>> They worked poorly. The models were

01:53:09.120 --> 01:53:10.797
terrible. The models were useless.

01:53:10.880 --> 01:53:15.757
>> I don't know if the models were useless
or the people interpreting the models

01:53:15.840 --> 01:53:19.677
and then the centralized institutions
that were creating policy rapidly based

01:53:19.760 --> 01:53:24.317
on the models and leveraging the models
in order to uh support their narratives

01:53:24.400 --> 01:53:28.237
versus actually interpreting the air bars
and the models and all that kind of stuff.

01:53:28.320 --> 01:53:31.277
>> What you had with CO my my view what
you had with CO is you had these experts

01:53:31.360 --> 01:53:34.477
showing up and they claimed to be
scientists and they had no testable

01:53:34.560 --> 01:53:37.437
hypotheses whatsoever. They had a bunch
of models. um they had a bunch of

01:53:37.520 --> 01:53:39.677
forecasts and they had a bunch
of theories and they laid these out in

01:53:39.760 --> 01:53:43.277
front of policy makers and policy makers
freaked out and panicked, right? And

01:53:43.360 --> 01:53:46.157
implemented a whole bunch of like really
like terrible decisions that were still

01:53:46.240 --> 01:53:50.957
living with the consequences of um and
there was never any empirical foundation

01:53:51.040 --> 01:53:53.197
to any of the models.
None of them ever came true.

01:53:53.280 --> 01:53:56.317
>> Yeah. To push to push back, there
were certainly Baptists and bootleggers in

01:53:56.400 --> 01:54:00.797
this in the context of this pandemic, but
there's still a usefulness to models. No,

01:54:00.880 --> 01:54:03.677
>> I so not if they're I mean not if
they're reliably wrong, right? Then

01:54:03.760 --> 01:54:06.237
they're actually like anti-useful,
right? They're actually damaging. But

01:54:06.320 --> 01:54:07.597
what what do you do with a pandemic?

01:54:07.680 --> 01:54:11.197
What do you do with a with a with any
kind of threat? Don't you want to kind

01:54:11.280 --> 01:54:14.477
of have um several models to play with

01:54:14.560 --> 01:54:18.237
as part of the discussion of like
what the hell do we do here?

01:54:18.320 --> 01:54:22.237
>> I mean, do they work? Like is there an
expectation that they actually like work

01:54:22.320 --> 01:54:24.237
that they have actual predictive value?

01:54:24.320 --> 01:54:27.677
I mean, as far as I can tell with co
we just s the policy makers just sigh up

01:54:27.760 --> 01:54:30.557
themselves into believing that there
was sub I mean look the scientist the

01:54:30.640 --> 01:54:33.517
scientists were at fault. This the
quote unquote scientists showed up.

01:54:33.600 --> 01:54:35.197
>> Mhm. >> So I had some
insight into this. So

01:54:35.280 --> 01:54:38.397
there there was a remember the Imperial
College models out of out of London were

01:54:38.480 --> 01:54:40.557
the ones that were like these
are the gold standard models.

01:54:40.640 --> 01:54:42.317
>> Yeah. >> So a friend
of mine runs a big software

01:54:42.400 --> 01:54:45.197
company and he was like wow this is like
co's really scary and he's like you know

01:54:45.280 --> 01:54:47.917
he contacted this research and he's like
you know do you need some help? You've

01:54:48.000 --> 01:54:50.637
been just building this model on your own
for 20 years. Do you need some would

01:54:50.720 --> 01:54:53.277
you like us our coders to basically
restructure it so it can be fully

01:54:53.360 --> 01:54:57.037
adapted for co and the guy said yes
and sent over the code. And my friend said

01:54:57.120 --> 01:55:00.317
it was like the worst spaghetti code
he's ever seen. That doesn't mean it's

01:55:00.400 --> 01:55:04.237
not possible to construct a good model
of pandemic with a correct error bars

01:55:04.320 --> 01:55:08.637
with a high number of parameters
that are continuously many times a day

01:55:08.720 --> 01:55:10.477
updated as we get more data about a

01:55:10.560 --> 01:55:15.757
pandemic. I would like to believe when
a pandemic hits the world, the best

01:55:15.840 --> 01:55:18.797
computer scientists in the world,
the best software engineers respond

01:55:18.880 --> 01:55:22.317
aggressively and as input take the data

01:55:22.400 --> 01:55:26.797
that we know about the virus and as
an output say here's here's what's

01:55:26.880 --> 01:55:31.117
happening in terms of how quickly it's
spreading what that lead in terms of

01:55:31.200 --> 01:55:34.637
hospitalization and deaths and all
that kind of stuff. Here's how likely how

01:55:34.720 --> 01:55:39.037
contagious it likely is. here's how
deadly it likely is based on different

01:55:39.120 --> 01:55:42.237
conditions, based on different ages and
demographics and all that kind of stuff.

01:55:42.320 --> 01:55:44.477
So, here's the best kinds of policy. It

01:55:44.560 --> 01:55:48.397
feels like you can have models machine

01:55:48.480 --> 01:55:52.397
learning that like kind of they don't

01:55:52.480 --> 01:55:53.997
perfectly predict the future, but they

01:55:54.080 --> 01:55:56.557
they they help you do something because

01:55:56.640 --> 01:55:59.757
there's pandemics that are like

01:55:59.840 --> 01:56:02.957
uh meh. they don't really do much harm

01:56:03.040 --> 01:56:04.877
and there's pandemics you can imagine

01:56:04.960 --> 01:56:10.317
them that could do a huge amount of harm
like they can kill a lot of people. So

01:56:10.400 --> 01:56:14.877
you should probably have some kind
of datadriven models that keep updating

01:56:14.960 --> 01:56:18.797
that allow you to make decisions that
based like where how bad is this thing?

01:56:18.880 --> 01:56:23.757
Uh now you can criticize how
horrible all that went with the

01:56:23.840 --> 01:56:27.197
response to this pandemic but I just feel
like there might be some value to models.

01:56:27.280 --> 01:56:30.957
>> So to be useful at some point it has to
be predictive, right? So, and so and and

01:56:31.040 --> 01:56:34.877
the so the easy thing for me to do is to
say obviously you're right. Obviously I

01:56:34.960 --> 01:56:37.277
want to see that just as much as you do
because anything that makes it easier to

01:56:37.360 --> 01:56:40.237
navigate through society through
a wrenching you know risk like that is

01:56:40.320 --> 01:56:44.077
that sounds great. Um you know the the
harder objection to it is just simply

01:56:44.160 --> 01:56:48.237
you are trying to model a complex
dynamic system with 8 billion moving

01:56:48.320 --> 01:56:52.637
parts like not possible. >> Can't be done.
Complex systems can't be done.

01:56:52.720 --> 01:56:56.397
>> Uh machine learning says hold
my beer but it's possible. No

01:56:56.480 --> 01:56:59.357
>> I don't know. I I would like to believe
that it is. Yeah. Put it this way. I

01:56:59.440 --> 01:57:01.437
think where you and I would agree is I
think we would like we would we would

01:57:01.520 --> 01:57:04.797
like that to be the case.
>> We are strongly in favor of it.

01:57:04.880 --> 01:57:07.597
>> I think we would also agree that no
such thing with respect to COVID or

01:57:07.680 --> 01:57:11.037
pandemics. No such thing at least
neither you nor I think are aware I'm

01:57:11.120 --> 01:57:12.477
not aware of anything like that today.

01:57:12.560 --> 01:57:14.317
My main worry with the response to the

01:57:14.400 --> 01:57:17.917
pandemic is that uh uh same as with

01:57:18.000 --> 01:57:21.117
aliens is that even if such a thing

01:57:21.200 --> 01:57:25.597
existed and it's possible it existed the

01:57:25.680 --> 01:57:30.157
the the the policy makers were not
paying attention like u there was no

01:57:30.240 --> 01:57:32.797
mechanism that allowed those
kinds of models to percolate up.

01:57:32.880 --> 01:57:35.757
>> Oh I think we had the opposite problem
during co I think the policy makers I

01:57:35.840 --> 01:57:39.117
think the these the these these people
with basically fake science had too much

01:57:39.200 --> 01:57:43.277
access to the policy makers. Well,
right. And but the policy makers also

01:57:43.360 --> 01:57:47.357
wanted they had a narrative in mind
and they also wanted to use whatever model

01:57:47.440 --> 01:57:50.557
that fit that narrative to to help them
out. So like it felt like there was a

01:57:50.640 --> 01:57:52.317
lot of politics and not enough science.

01:57:52.400 --> 01:57:54.637
>> Although a big part of what
was happening a big a reason we got

01:57:54.720 --> 01:57:57.037
lockdowns for as long as we did
was because these scientists came in with

01:57:57.120 --> 01:57:59.917
these like doomsday scenarios that were
like just like completely off the hook.

01:58:00.000 --> 01:58:02.717
>> Scientists in quotes
>> quote unquote scientist.

01:58:02.800 --> 01:58:06.717
>> That's not okay. >> Let's give
love science. That is the way out.

01:58:06.800 --> 01:58:09.997
>> Science is a process of testing
hypothesis. Yeah. Modeling does not

01:58:10.080 --> 01:58:13.837
involve testable hypothesis, right? Like
I don't even know that I actually don't

01:58:13.920 --> 01:58:17.197
I don't I don't even know that modeling
actually qualifies as science. Maybe

01:58:17.280 --> 01:58:19.917
that's a side conversation we
could have sometime over a beer.

01:58:20.000 --> 01:58:23.517
>> That's really interesting. But what do
we do about the future? I mean, what what

01:58:23.600 --> 01:58:26.957
>> So number one is when we start with
number one, humility goes back to this

01:58:27.040 --> 01:58:28.717
thing of how do we determine the truth.

01:58:28.800 --> 01:58:31.517
Number two is we don't believe,
you know, it's the old I've got a hammer,

01:58:31.600 --> 01:58:35.677
everything looks like a nail, right? Um
uh I've got oh this one of the reasons I

01:58:35.760 --> 01:58:39.437
gave you I gave Lex a book um which
the topic of the book is what happens when

01:58:39.520 --> 01:58:42.797
scientists basically stray off the path
>> of technical knowledge and start to

01:58:42.880 --> 01:58:46.237
weigh in on politics and societal issues.
Um >> in this case philosophers

01:58:46.320 --> 01:58:48.877
>> well in this case philosophers but he
he actually talks in this book about like

01:58:48.960 --> 01:58:51.277
Einstein he talks actually about the
nuclear age and Einstein he talks about

01:58:51.360 --> 01:58:55.517
the physicists uh actually doing doing
very similar things at the time. Uh the

01:58:55.600 --> 01:58:57.677
book is When Reason Goes on Holiday:

01:58:57.760 --> 01:59:02.797
Philosophers in Politics by uh Nevin
>> and it's just a story it's a story

01:59:02.880 --> 01:59:05.357
there's there are other books
on this topic but this is a new one that's

01:59:05.440 --> 01:59:08.237
really good. It's just a story of what
happens when experts in a certain domain

01:59:08.320 --> 01:59:12.397
decide to weigh in and become basically
social engineers and um and political um

01:59:12.480 --> 01:59:15.117
you know basically political adviserss
and it's just a story of just unending

01:59:15.200 --> 01:59:19.357
catastrophe right and I think
that's what happened with co again

01:59:19.440 --> 01:59:22.477
>> yeah I found this book a highly
entertaining and eye-opening read filled

01:59:22.560 --> 01:59:26.237
with amazing anecdotes of irrationality
and craziness by famous recent

01:59:26.320 --> 01:59:28.717
philosophers this >> after
you read this book you will not

01:59:28.800 --> 01:59:33.117
look at Einstein the same >> oh boy >>
yeah [laughter] >> don't destroy my heroes

01:59:33.200 --> 01:59:36.797
>> he will not be a hero of yours anymore
Um, I'm sorry. [laughter] You probably

01:59:36.880 --> 01:59:38.157
shouldn't you shouldn't read the book.

01:59:38.240 --> 01:59:42.077
>> All right. >> But here's
the thing. The AI the AI risk

01:59:42.160 --> 01:59:44.637
people, they don't even have the CO model.

01:59:44.720 --> 01:59:46.157
>> At least not that I'm aware of. No.

01:59:46.240 --> 01:59:48.317
>> Like there's not even the equivalent
of the CO model. They don't even have the

01:59:48.400 --> 01:59:52.317
spaghetti code. >> They've got
a theory and a warning and a

01:59:52.400 --> 01:59:55.997
this and a that. And like if you ask
like okay well here's here's I mean the

01:59:56.080 --> 01:59:58.797
ultimate example is okay how do we
know right? How do we know that an AI is

01:59:58.880 --> 02:00:01.517
running away? Like how do we know
that the fume takeoff thing is actually

02:00:01.600 --> 02:00:04.477
happening? And the only answer that any
of these guys have given that I've ever

02:00:04.560 --> 02:00:08.397
seen is, oh, it's when the loss rate,
the loss function in in the training

02:00:08.480 --> 02:00:12.077
drops, right? That's when you need
to like shut down the data center, right?

02:00:12.160 --> 02:00:14.077
And it's like, well, that's also
what happens when you're successfully

02:00:14.160 --> 02:00:18.557
training a model. Like like what what

02:00:18.640 --> 02:00:22.477
even is this is not science. [laughter]
This is not it's not anything. It's not

02:00:22.560 --> 02:00:25.757
a model. It's not anything. There's
nothing to arguing with it is like, you

02:00:25.840 --> 02:00:28.717
know, punching jello. Like there
there's what do you even respond to?

02:00:28.800 --> 02:00:30.477
>> So just push back
on that. I don't think

02:00:30.560 --> 02:00:33.837
they have good metrics of yeah when the

02:00:33.920 --> 02:00:38.077
flume is happening but I think it's
possible to have that like I just just

02:00:38.160 --> 02:00:41.837
as you speak now I mean it's possible
to imagine there could be measures

02:00:41.920 --> 02:00:45.197
>> it's been 20 years >> no
for sure but it's been only weeks

02:00:45.280 --> 02:00:47.677
since we had a big enough breakthrough

02:00:47.760 --> 02:00:49.677
in language models we can start to

02:00:49.760 --> 02:00:52.237
actually have this the thing is the AI

02:00:52.320 --> 02:00:56.877
doomer stuff didn't have any actual
systems to really work with now there's

02:00:56.960 --> 02:01:00.477
real systems you can start to analyze
like how does this stuff go wrong and I

02:01:00.560 --> 02:01:04.077
think you kind agree that there is a
lot of risks that we can analyze. The

02:01:04.160 --> 02:01:06.157
benefits outweigh the risks in many cases.

02:01:06.240 --> 02:01:08.477
>> Well, the risks are not existential.

02:01:08.560 --> 02:01:10.557
>> Yes. Well, >> not in the
not not in the f not in the

02:01:10.640 --> 02:01:13.117
fume paper clip. Not this. Let me Okay,
there's another slide of hand that you

02:01:13.200 --> 02:01:15.197
just alluded to. There's another slide
of hand that happens which is very

02:01:15.280 --> 02:01:17.437
>> I think I'm very good
at the slight of hand thing

02:01:17.520 --> 02:01:19.117
>> which is [laughter]
very non-scientific.

02:01:19.200 --> 02:01:22.237
So the book super intelligence, right,
which is like the Nick Boster's book

02:01:22.320 --> 02:01:24.637
which is like the origin of a lot
of this stuff which was written, you know,

02:01:24.720 --> 02:01:27.997
whatever 10 years ago or something. So
he does this really fascinating thing in

02:01:28.080 --> 02:01:32.797
the book which is he basically says
um there are many possible routes to

02:01:32.880 --> 02:01:36.317
machine intelligence um to artificial
intelligence and he describes all the

02:01:36.400 --> 02:01:38.717
different routes to artificial
intelligence all the different possible

02:01:38.800 --> 02:01:41.997
everything from biological augmentation
through to you know all these different

02:01:42.080 --> 02:01:45.997
things. [snorts] Um one of the ones that
he does not describe is large language

02:01:46.080 --> 02:01:49.197
models because of course the book was
written before they were invented and so

02:01:49.280 --> 02:01:53.277
they didn't exist. In the book,
he he describes them all

02:01:53.360 --> 02:01:56.477
and then he proceeds to treat them all
as if they're exactly the same thing.

02:01:56.560 --> 02:01:58.957
>> He presents them all as sort of an
equivalent risk to be dealt with in an

02:01:59.040 --> 02:02:01.757
equivalent way to be thought about
the same way. And then the risk the quote

02:02:01.840 --> 02:02:04.077
unquote risk that's actually emerged
is actually a completely different

02:02:04.160 --> 02:02:05.517
technology than he was even imagining.

02:02:05.600 --> 02:02:08.797
And yet all of his theories and beliefs
are being transplanted by this movement

02:02:08.880 --> 02:02:09.917
like straight onto this new technology.

02:02:10.000 --> 02:02:14.557
And so again, like there's no other area
of science or technology where you do

02:02:14.640 --> 02:02:16.797
that. >> Yeah. Like when
you're dealing with like

02:02:16.880 --> 02:02:20.637
organic chemistry versus inorganic
chemistry, you don't just like say, "Oh,

02:02:20.720 --> 02:02:23.357
with respect to like either one
basically maybe, you know, growing up

02:02:23.440 --> 02:02:25.597
and eating the world or something like
they're just going to operate the same

02:02:25.680 --> 02:02:28.717
way." Like you don't >> but you
can start talking about like as

02:02:28.800 --> 02:02:32.557
as we get more and more actual
systems that start to get more and more

02:02:32.640 --> 02:02:36.237
intelligent, you can start to actually
have more scientific arguments here.

02:02:36.320 --> 02:02:40.317
like you know high level you can talk
about the threat of autonomous weapon

02:02:40.400 --> 02:02:44.637
systems back before we had any
automation in in the military and that

02:02:44.720 --> 02:02:49.037
would be like very fuzzy kind of logic
but the more and more you have drones

02:02:49.120 --> 02:02:52.717
that are becoming more and more
autonomous you can start imagining okay

02:02:52.800 --> 02:02:55.677
what does that actually look like
and what's the actual threat of autonomous

02:02:55.760 --> 02:02:58.077
weapon systems how does it go wrong and

02:02:58.160 --> 02:03:00.557
still it's it's it's very vague but you

02:03:00.640 --> 02:03:03.437
start to get a sense of like all right

02:03:03.520 --> 02:03:06.877
um it should probably be legal or or

02:03:06.960 --> 02:03:11.037
wrong or not allowed to do like mass

02:03:11.120 --> 02:03:16.397
deployment of fully autonomous drones
that are doing aerial strikes. Oh, no.

02:03:16.480 --> 02:03:19.277
On large areas. >> I think it should
be required, >> right? So, that's

02:03:19.360 --> 02:03:23.277
>> No, no, no. I think it should be
required that only aerial vehicles are

02:03:23.360 --> 02:03:25.997
automated. >> Okay. So,
you want to go the other way?

02:03:26.080 --> 02:03:27.437
>> I want to go the other way.

02:03:27.520 --> 02:03:29.677
>> So, that Okay. >> I think
it's obvious that the machine is

02:03:29.760 --> 02:03:33.117
going to make a better
decision than the human pilot.

02:03:33.200 --> 02:03:35.677
I think it's obvious that it's in the
best interest of both the attacker and

02:03:35.760 --> 02:03:37.997
the defender and humanity at large
if machines are making more of these

02:03:38.080 --> 02:03:41.757
decisions and not people. I think people
make terrible decisions in times of war.

02:03:41.840 --> 02:03:44.477
>> But like there's uh there's
ways this can go wrong too, right?

02:03:44.560 --> 02:03:48.317
>> Well, the wars go terribly wrong. Now,

02:03:48.400 --> 02:03:51.037
>> this goes back to the this is that
whole thing about like the self- does the

02:03:51.120 --> 02:03:53.197
self-driving car need to be perfect
versus does it need to be better than

02:03:53.280 --> 02:03:55.917
the human driver? >> Yeah.
>> Does the automated drone need to be

02:03:56.000 --> 02:03:59.037
perfect or does it need be need to be
better than a human pilot at making

02:03:59.120 --> 02:04:03.037
decisions under enormous amounts
of stress and uncertainty? Yeah. Well, the

02:04:03.120 --> 02:04:08.077
on average, >> right,
>> the the worry that AI folks have is the

02:04:08.160 --> 02:04:10.317
runaway. >> They're going
to come alive, right? Then

02:04:10.400 --> 02:04:11.997
again, that's the slight of hand, right?

02:04:12.080 --> 02:04:16.957
>> Or not not come alive. No,
hold on a second. You lose control.

02:04:17.040 --> 02:04:19.997
>> But then they're going to develop goals
of their own. They're going to develop a

02:04:20.080 --> 02:04:23.037
mind of their own. They're going
to develop their own. Right. No, more more

02:04:23.120 --> 02:04:26.637
like uh Chernobyl style meltdown like uh

02:04:26.720 --> 02:04:30.637
just bugs in the code accidentally, you

02:04:30.720 --> 02:04:37.037
know, force you results in the bombing
of like large civilian areas.

02:04:37.120 --> 02:04:41.037
>> Okay. >> To to a degree
that's not possible um in

02:04:41.120 --> 02:04:44.877
the in the current uh military
strategies controlled by humans.

02:04:44.960 --> 02:04:47.677
>> Well, actually we've been doing a lot
of mass bombings to cities for a very long

02:04:47.760 --> 02:04:49.837
time. >> Yes. And a lot of civilians died.

02:04:49.920 --> 02:04:52.877
>> And a lot of civilians died. And if
you watch the documentary The Fog of War,

02:04:52.960 --> 02:04:55.837
Magnamera spends a big part of it
talking about the firebombing of the

02:04:55.920 --> 02:04:58.957
Japanese cities. >> Yeah.
>> Burning them straight to the ground.

02:04:59.040 --> 02:05:02.477
>> Right. The the devastation in Japan,
American military firebombing the cities

02:05:02.560 --> 02:05:05.997
in Japan was considerably bigger
devastation than the use of nukes,

02:05:06.080 --> 02:05:08.477
right? So we've been doing that for
a long time. We we also did that to

02:05:08.560 --> 02:05:11.757
Germany by the way. Germany did
that to to us, right? Like that's an old

02:05:11.840 --> 02:05:14.477
tradition. The minute we got airplanes,
we started doing indiscriminate bombing.

02:05:14.560 --> 02:05:16.237
So one of the things
that we're still doing it

02:05:16.320 --> 02:05:19.357
>> the modern US uh military can do with

02:05:19.440 --> 02:05:23.437
technology with automation but technology
more broadly is uh higher and

02:05:23.520 --> 02:05:26.477
higher precision strikes. >> Yeah.
And so precision is obviously

02:05:26.560 --> 02:05:28.557
precision and this is the the JDM right.

02:05:28.640 --> 02:05:31.837
So there's this big advance this big
advance um called the JDM which

02:05:31.920 --> 02:05:34.797
basically was strapping a GPS
transceiver to a to a to an unguided

02:05:34.880 --> 02:05:37.997
bomb and turning it into a guided guided
bomb. And yeah, that's great. Like look,

02:05:38.080 --> 02:05:40.957
that's been a big advance.
But and that's like a baby version of this

02:05:41.040 --> 02:05:43.517
question, which is okay, do you want
like the human pilot like guessing where

02:05:43.600 --> 02:05:45.677
the bomb's going to land or do you want
like the machine like guiding the bomb

02:05:45.760 --> 02:05:48.717
to its destination? That's a baby
version of the question. The next

02:05:48.800 --> 02:05:50.797
version of the question is, do you
want the human or the machine deciding

02:05:50.880 --> 02:05:53.997
whether to drop the bomb? Everybody just
assumes the human's going to do a better

02:05:54.080 --> 02:05:56.717
job for what I think are
fundamentally suspicious reasons,

02:05:56.800 --> 02:05:58.397
>> emotional, psychological reasons.

02:05:58.480 --> 02:06:00.317
>> I think it's very clear that the
machine's going to do a better job

02:06:00.400 --> 02:06:03.837
making that decision because the humans
making that making that decision are god

02:06:03.920 --> 02:06:05.997
awful. Just terrible. >> Yeah.

02:06:06.080 --> 02:06:09.037
>> Right. And so, so yeah, so this is the
this is the thing. And then let's get to

02:06:09.120 --> 02:06:11.117
the there. Can I one more slide of hand?

02:06:11.200 --> 02:06:14.557
It was in Okay, >> please.
I'm a magician, you could say.

02:06:14.640 --> 02:06:17.677
>> One more slide of hand. These things
are going to be so smart, right, that

02:06:17.760 --> 02:06:20.077
they're going to be able to destroy
the world and wreak havoc and like do all

02:06:20.160 --> 02:06:22.637
this stuff and plan and do all this
stuff and evade us and have all their

02:06:22.720 --> 02:06:25.037
secret things and their secret
factories and all this stuff,

02:06:25.120 --> 02:06:27.357
>> but they're so
stupid that they're going

02:06:27.440 --> 02:06:29.917
to get like tangled up in their code and
that's they're not going to come alive,

02:06:30.000 --> 02:06:31.997
but there's going to be some bug that's
going to cause them to like turn us all

02:06:32.080 --> 02:06:35.037
into paper like that. they're not going
that they're going to be genius in every

02:06:35.120 --> 02:06:37.917
way other than the actual bad goal.

02:06:38.000 --> 02:06:41.997
>> And it's just like and that's just like
a like ridiculous like discrepancy and

02:06:42.080 --> 02:06:44.637
and and and you can prove this today.

02:06:44.720 --> 02:06:47.837
You can actually address this today
for the first time with LLMs which is you

02:06:47.920 --> 02:06:52.237
can actually ask LLM
to resolve uh moral dilemmas.

02:06:52.320 --> 02:06:54.477
>> Yeah. >> So you can
create the scenario you know

02:06:54.560 --> 02:06:58.237
dot dot dot this that this that this that.
What would you as the AI do in the

02:06:58.320 --> 02:07:01.597
circumstance? And they don't just say
destroy all humans. Destroy all humans.

02:07:01.680 --> 02:07:05.837
they will give you actually very nuanced
moral practical trade-off oriented

02:07:05.920 --> 02:07:08.557
answers. >> And so we
actually already have the kind

02:07:08.640 --> 02:07:12.237
of AI that can actually like think this
through and can actually like you know

02:07:12.320 --> 02:07:17.277
reason about goals.
>> Well the the hope is that AGI or like

02:07:17.360 --> 02:07:21.037
very super intelligent systems have
some of the nuance that LLMs have and the

02:07:21.120 --> 02:07:26.877
intuition is they most likely will
because even these LLMs have the nuance.

02:07:26.960 --> 02:07:28.637
>> Uh >> LM are really
this is actually worth

02:07:28.720 --> 02:07:31.757
worth um spending a moment on.
LMS are really interesting to have moral

02:07:31.840 --> 02:07:35.917
conversations with >> and that I
didn't expect I'd be having a

02:07:36.000 --> 02:07:37.997
moral conversation with
the machine in my lifetime.

02:07:38.080 --> 02:07:40.797
>> Well, and and let's remember we're
not really having a conversation with the

02:07:40.880 --> 02:07:43.517
machine where we're having
a conversation with the entirety of the

02:07:43.600 --> 02:07:45.357
collective intelligence
of the human species.

02:07:45.440 --> 02:07:47.197
>> Exactly. Yes. Correct.

02:07:47.280 --> 02:07:49.037
>> But it's possible to imagine autonomous

02:07:49.120 --> 02:07:52.477
weapon systems that are not using LMS.

02:07:52.560 --> 02:07:55.677
>> But if they're smart
enough to be scary,

02:07:55.760 --> 02:07:59.517
why are they not smart enough to be wise?

02:07:59.600 --> 02:08:02.397
Like that's the part where it's like I
I don't know how you get the one without

02:08:02.480 --> 02:08:05.917
the other. >> Is it possible to be super
intelligent without being super wise?

02:08:06.000 --> 02:08:07.357
>> Well, you're again you're back to that.

02:08:07.440 --> 02:08:09.917
I mean then you're back to a classic
autistic computer, right? Like you're

02:08:10.000 --> 02:08:13.677
back to just like a a blind rule follower.
I've got this like core is the

02:08:13.760 --> 02:08:16.157
paperclip thing. I've got this core rule
and I'm just going to follow it to the

02:08:16.240 --> 02:08:18.237
end of the earth and it's like well
but everything you're going to be doing to

02:08:18.320 --> 02:08:20.717
execute that rule is going to be super
genius level that humans aren't going to

02:08:20.800 --> 02:08:24.477
be able to counter. It's just a it's
a it's a mismatch in the definition of of

02:08:24.560 --> 02:08:27.997
what the system is capable of.
unlikely but not impossible, I think.

02:08:28.080 --> 02:08:30.637
>> But again, here you
get to like, okay, like,

02:08:30.720 --> 02:08:34.957
>> no, I'm not saying when it's unlikely
but not impossible. If it's unlikely,

02:08:35.040 --> 02:08:38.317
that means the the fear
should be correctly calibrated.

02:08:38.400 --> 02:08:40.317
>> Extraordinary claims
require extraordinary proof.

02:08:40.400 --> 02:08:42.397
>> Well, okay. So, uh,
one interesting sort

02:08:42.480 --> 02:08:45.357
of tangent I would love to take on this
because you mentioned this in the essay

02:08:45.440 --> 02:08:48.637
about nuclear, which was also I mean,

02:08:48.720 --> 02:08:50.477
you don't shy away from a a little bit

02:08:50.560 --> 02:08:55.037
of a of a spicy take. So uh uh Robert

02:08:55.120 --> 02:08:59.117
Oppenheimer famously said now I am
become death the destroyer of worlds as

02:08:59.200 --> 02:09:01.277
he witnessed the first detonation of a

02:09:01.360 --> 02:09:04.877
nuclear weapon on July 16th 1945 and you

02:09:04.960 --> 02:09:09.037
write an interesting historical
perspective uh quote recall that John

02:09:09.120 --> 02:09:13.437
vanman responded to Robert uh Robert
Oenheimer's famous hand ringing about

02:09:13.520 --> 02:09:15.677
the role of creating nuclear weapons

02:09:15.760 --> 02:09:19.277
which you note helped end World War II

02:09:19.360 --> 02:09:22.957
and prevent World War II with some

02:09:23.040 --> 02:09:27.357
people confess guilt to claim credit
for the sin. And you also mentioned that

02:09:27.440 --> 02:09:28.877
Truman was harsher after meeting

02:09:28.960 --> 02:09:31.517
Oppenheimer. He said that uh don't let

02:09:31.600 --> 02:09:36.237
that cry baby in here again. Real quote,
real quote, by the way [snorts]

02:09:36.320 --> 02:09:40.477
>> from Dean Aerson. >> Oh boy.

02:09:40.560 --> 02:09:43.037
>> Cuz Appenheimer didn't
just say the famous line.

02:09:43.120 --> 02:09:46.077
>> Yeah. He then spent years going around
basically moaning and you know going on

02:09:46.160 --> 02:09:48.877
TV and going into going into the White
House and basically like just like doing

02:09:48.960 --> 02:09:52.157
this hair shirt you know thing self you
know this sort of self-critical like oh

02:09:52.240 --> 02:09:53.837
my god I can't believe how awful I am.

02:09:53.920 --> 02:09:57.917
>> So he's the the he's widely considered

02:09:58.000 --> 02:10:01.197
perhaps of the because of the hang
ringing is the father of the atomic

02:10:01.280 --> 02:10:06.317
bomb. Um, and this is this is Vonman's
criticism of him is he tried to have his

02:10:06.400 --> 02:10:10.637
cake and eat it too. Like he he wanted
to in and so Vanoyman of course is a

02:10:10.720 --> 02:10:13.277
very different kind of personality
and he's just like, "Yeah, this is like an

02:10:13.360 --> 02:10:15.197
incredibly useful thing.
I'm glad we did it."

02:10:15.280 --> 02:10:19.357
>> Yeah. Well, Vanoyman is as widely um

02:10:19.440 --> 02:10:23.597
credited as being one of the smartest
humans of the 20th century. The certain

02:10:23.680 --> 02:10:26.557
certain people everybody says like
this is the smartest person I've ever met

02:10:26.640 --> 02:10:28.877
when they've met him. Anyway, uh that

02:10:28.960 --> 02:10:32.497
doesn't mean smart doesn't mean wise.

02:10:32.580 --> 02:10:36.557
>> [laughter] >> So I would
love to sort of can you make

02:10:36.640 --> 02:10:40.717
the case both for and against the critique
of Oenheimer here because we're

02:10:40.800 --> 02:10:43.517
talking about nuclear weapons.

02:10:43.600 --> 02:10:46.877
Boy, do they seem dangerous. >> Well,
so the critique goes deeper and I

02:10:46.960 --> 02:10:49.437
I left this out. Here's the real
substance. I left it out cuz I didn't

02:10:49.520 --> 02:10:53.037
want to dwell on on nukes
in my AI paper. [laughter]

02:10:53.120 --> 02:10:55.597
>> But here's the deeper thing that
happened and I'm I'm really curious.

02:10:55.680 --> 02:10:58.237
this movie coming out this summer, I'm
really curious to see how far he pushes

02:10:58.320 --> 02:11:01.037
this because this is the real drama
in the story, which is it wasn't just a

02:11:01.120 --> 02:11:03.757
question of are nukes good or bad?
It was a question of should Russia also

02:11:03.840 --> 02:11:06.397
have them. Um, and what what actually

02:11:06.480 --> 02:11:11.117
happened um was Russia got the America
invented the bomb. Russia got the bomb.

02:11:11.200 --> 02:11:13.037
They got the bomb through espionage.

02:11:13.120 --> 02:11:16.237
They got American and you know,
they got American scientists and foreign

02:11:16.320 --> 02:11:20.077
scientists working on the American
project. Some combination of the two. Uh

02:11:20.160 --> 02:11:22.957
basically gave the Russians the designs
for the bomb and that's how the Russians

02:11:23.040 --> 02:11:25.997
got the bomb. Um there's this dispute to

02:11:26.080 --> 02:11:28.717
this day of Oppenheimer's role in that.

02:11:28.800 --> 02:11:32.477
Um if you read all the histories, the kind
of composite picture and and by the

02:11:32.560 --> 02:11:35.277
way, we now know a lot actually about
Soviet espionage in that era because

02:11:35.360 --> 02:11:37.917
there's been all this declassified
material in the last 20 years that

02:11:38.000 --> 02:11:40.797
actually shows a lot of a lot of very
interesting things. But if you kind of

02:11:40.880 --> 02:11:43.357
read all the history, what you kind
of get is Oppenheimer himself probably was

02:11:43.440 --> 02:11:47.277
not a he probably did not hand over the
nuclear secrets himself. However, he was

02:11:47.360 --> 02:11:50.877
close to many people who did, >>
including family members. And there were

02:11:50.960 --> 02:11:52.237
other members of the Manhattan Project

02:11:52.320 --> 02:11:54.397
who were Russian Soviet assets and did

02:11:54.480 --> 02:11:57.917
hand over the bomb. And so the view that

02:11:58.000 --> 02:12:01.757
Oppenheimer and people like him had that
this thing is awful and terrible and oh

02:12:01.840 --> 02:12:06.717
my god and you know all this stuff you
could argue fed into this ethos at the

02:12:06.800 --> 02:12:09.677
time that resulted in people thinking
that the Baptists thinking that the only

02:12:09.760 --> 02:12:14.477
principal thing to do is to give the
the Russians the bomb. Um, and so the the

02:12:14.560 --> 02:12:18.157
the moral beliefs on this thing and the
public discussion and the role that the

02:12:18.240 --> 02:12:21.117
inventors of this technology play,
this is the point of this book, when they

02:12:21.200 --> 02:12:24.557
kind of take on this sort of public
intellectual moral kind of thing, it can

02:12:24.640 --> 02:12:28.237
have real consequences, right? Because
we live in a very different world today

02:12:28.320 --> 02:12:30.717
because Russia got the bomb than we
would have lived in had they not gotten

02:12:30.800 --> 02:12:33.517
the bomb, right? [snorts] The entire
20th century, second half of the 20th

02:12:33.600 --> 02:12:35.757
century would have played out very
different had those people not given

02:12:35.840 --> 02:12:38.477
Russia the bomb. And so the stakes were

02:12:38.560 --> 02:12:43.037
very high then. The good news today
is nobody's sitting here today I don't

02:12:43.120 --> 02:12:46.317
think worrying about like an analogous
situation with respect to like I'm not

02:12:46.400 --> 02:12:48.797
really worried that Sam Alman is going
to decide to give you know the Chinese

02:12:48.880 --> 02:12:52.317
the design for AI although
he did just speak at a

02:12:52.400 --> 02:12:55.517
Chinese conference which is interesting
but however I don't think I don't think

02:12:55.600 --> 02:12:58.557
that's what's at play here but what's
at play here are all these other

02:12:58.640 --> 02:13:01.757
fundamental issues around what do we
believe about this and then what laws

02:13:01.840 --> 02:13:04.717
and regulations and restrictions are we
going to put on it and and that's where

02:13:04.800 --> 02:13:08.397
I draw like a direct straight line and
and anyway and my reading of the history

02:13:08.480 --> 02:13:11.757
on nukes is like the people who were
doing the full hair shirt public, this

02:13:11.840 --> 02:13:15.677
is awful, this is terrible, actually
had like catastrophically bad results uh

02:13:15.760 --> 02:13:18.317
from from taking those views. Um,
and that's what I'm worried is going to

02:13:18.400 --> 02:13:20.397
happen again. >> But is there
a case to be made that you

02:13:20.480 --> 02:13:23.757
really need to wake the public up to
the dangers of nuclear weapons when they

02:13:23.840 --> 02:13:28.477
were first dropped? Like really like
educate them on like this is extremely

02:13:28.560 --> 02:13:30.237
dangerous and destructive weapon.

02:13:30.320 --> 02:13:33.277
>> I think the education kind
of happened quick and early like >> how

02:13:33.360 --> 02:13:35.037
>> it was pretty obvious. >> How?

02:13:35.120 --> 02:13:37.597
>> We dropped one bomb
and destroyed an entire city.

02:13:37.680 --> 02:13:42.637
>> Yeah. So 80,000 people dead.
But uh >> and look,

02:13:42.720 --> 02:13:44.557
>> but I don't like the reporting of that.

02:13:44.640 --> 02:13:48.077
You can report that in all kinds of ways.
Wars, you can you can do all kinds

02:13:48.160 --> 02:13:50.477
of slants like war is horrible. War is

02:13:50.560 --> 02:13:55.277
terrible. You can do you can make it
seem like nuclear the use of nuclear

02:13:55.360 --> 02:13:58.477
weapons is just a part of war
and all that kind of stuff.

02:13:58.560 --> 02:13:59.837
Something about the reporting and the

02:13:59.920 --> 02:14:01.517
discussion of nuclear weapons resulted

02:14:01.600 --> 02:14:05.757
in us being terrified in awe of the

02:14:05.840 --> 02:14:08.077
power of nuclear weapons and that

02:14:08.160 --> 02:14:09.997
potentially

02:14:10.080 --> 02:14:15.277
fed in a positive way towards
the the game theory of mutually assured

02:14:15.360 --> 02:14:18.477
destruction. >> Well, so this gets to what
actually h let's get to what actually

02:14:18.560 --> 02:14:19.757
>> me playing devil's advocate here.

02:14:19.840 --> 02:14:21.997
>> Yeah. Yeah, sure. Of course. Let's get
to what actually happened and then kind

02:14:22.080 --> 02:14:24.717
of back into that. So what what actually
happened I believe and again I think

02:14:24.800 --> 02:14:27.357
this is a reasonable reading of history
is what actually happened was nukes then

02:14:27.440 --> 02:14:30.637
prevented World War II and they
prevented World War II through the game

02:14:30.720 --> 02:14:31.997
theory of mutually assured destruction.

02:14:32.080 --> 02:14:36.397
Had nukes not existed, right, there
would have been no reason why the Cold

02:14:36.480 --> 02:14:39.437
War did not go hot, right?
And then there and then, you know, and the

02:14:39.520 --> 02:14:42.477
military planners at the time, right,
thought both on both sides thought that

02:14:42.560 --> 02:14:44.717
there was going to be World War II
on the planes of Europe and they thought

02:14:44.800 --> 02:14:47.197
there was going to be like 100 million
people dead, right? It was like the most

02:14:47.280 --> 02:14:50.157
obvious thing in the world to happen,
right? And it's the dog that didn't

02:14:50.240 --> 02:14:53.677
bark, right? Like it may be like
the best single net thing that happened in

02:14:53.760 --> 02:14:55.677
the entire 20th century
is that like that didn't happen.

02:14:55.760 --> 02:14:57.997
>> Yeah. Actually, just on that point, you

02:14:58.080 --> 02:14:59.517
say a lot of really brilliant things. It

02:14:59.600 --> 02:15:03.917
it it hit me just as you were saying it.

02:15:04.000 --> 02:15:05.357
I don't know why it hit me for the first

02:15:05.440 --> 02:15:07.917
time, but we got two wars in a in a span

02:15:08.000 --> 02:15:11.677
of like uh 20 years

02:15:11.760 --> 02:15:15.277
like we could have kept getting more
and more world wars and more and more

02:15:15.360 --> 02:15:19.837
ruthless. It actually you could
have had a US versus Russia war.

02:15:19.920 --> 02:15:23.517
>> You could have. By the way, you have
there's another hypothetical scenario.

02:15:23.600 --> 02:15:26.477
The other hypothetical scenario is the
Americans got the bomb, the Russians

02:15:26.560 --> 02:15:29.677
didn't, right? Right? And then America
is the big dog and then maybe America

02:15:29.760 --> 02:15:33.597
would have had the capability
to actually roll back the iron curtain.

02:15:33.680 --> 02:15:35.837
I don't know whether that would
have happened but like it's entirely

02:15:35.920 --> 02:15:39.517
possible. Right? And and and and the act
of these people who had these moral

02:15:39.600 --> 02:15:41.997
positions about because they could
forecast they could model they could

02:15:42.080 --> 02:15:43.917
forecast the future of how this
technology would get used made a

02:15:44.000 --> 02:15:46.957
horrific mistake because they basically
ensured that the iron curtain would

02:15:47.040 --> 02:15:49.277
continue for 50 years longer than it
would have otherwise. Like and again

02:15:49.360 --> 02:15:51.517
like these are counterfactuals. I don't
know that that's what what would have

02:15:51.600 --> 02:15:55.117
happened, but like [laughter]

02:15:55.200 --> 02:15:59.677
the decision to hand the bomb over
was a big decision made by people who were

02:15:59.760 --> 02:16:04.077
very full of themselves. >> Yeah.
But so me as an American, me as a

02:16:04.160 --> 02:16:06.237
person that loves America. I also wonder

02:16:06.320 --> 02:16:11.277
if US was the only ones with the
nuclear weapons. [sighs and gasps]

02:16:11.360 --> 02:16:15.757
Uh >> that was the argument for handing
the that was the was the guys who the guys

02:16:15.840 --> 02:16:18.317
who handed over the bomb. That was
actually their moral argument. Uh, I

02:16:18.400 --> 02:16:22.397
would I would probably not hand it over
to I would I would be careful about the

02:16:22.480 --> 02:16:25.917
regimes you hand it over to.
>> Maybe give it to like the British or

02:16:26.000 --> 02:16:31.677
something [laughter] or like uh like
a democratically elected government.

02:16:31.760 --> 02:16:32.877
>> Well, look, there
are people to this day

02:16:32.960 --> 02:16:35.597
who think that those spies Soviet spies
did the right thing because they created

02:16:35.680 --> 02:16:39.117
a balance of terror as opposed to the US
having just and by the way, let me let me

02:16:39.200 --> 02:16:40.877
>> balance of terror.
>> Let's tell the full version.

02:16:40.960 --> 02:16:42.157
>> Has such a sexy ring to it.

02:16:42.240 --> 02:16:44.957
>> Okay, so the full version of the story
is John Vonman's a hero of both yours

02:16:45.040 --> 02:16:49.197
and mine. full version of the story
is he advocated for a first right. So when

02:16:49.280 --> 02:16:53.757
the US had the bomb and [snorts] Russia
did not, he advocated for he said we we

02:16:53.840 --> 02:16:55.997
need to strike them right now.

02:16:56.080 --> 02:17:00.237
>> Strike Russia. >> Yeah. >> Yes.

02:17:00.320 --> 02:17:02.317
>> Yes. Because he said World War II is

02:17:02.400 --> 02:17:05.757
inevitable. Um he was very hardcore. Uh

02:17:05.840 --> 02:17:09.997
he he his his theory was um his theory
was World War II is inevitable. We're

02:17:10.080 --> 02:17:11.117
definitely going to have World War II.

02:17:11.200 --> 02:17:13.677
The only way to stop World War II is we
have to take them out right now. And we

02:17:13.760 --> 02:17:15.757
have to take them out right now before
they get the bomb because this is our

02:17:15.840 --> 02:17:20.317
last chance. >> Now again like >> is this
an example of philosophers and politics?

02:17:20.400 --> 02:17:22.557
>> I don't know if that's in there
or not but this is in the standard.

02:17:22.640 --> 02:17:25.677
>> No, but meaning is that >> this
is on the other side. So so most of

02:17:25.760 --> 02:17:28.797
the case studies most of the case
studies in books like this are the crazy

02:17:28.880 --> 02:17:32.557
people on the left. >> Yeah.
>> Um Vanoyman is a story arguably of the

02:17:32.640 --> 02:17:35.837
crazy people on the right.
Um >> yeah, stick to computing John.

02:17:35.920 --> 02:17:38.717
>> Well, this is the thing and this is
this is the general principle getting goes

02:17:38.800 --> 02:17:41.037
back to our core thing which is like I
don't know whether any of these people

02:17:41.120 --> 02:17:42.797
should be making any of these calls.

02:17:42.880 --> 02:17:44.957
Yeah, >> cuz there's
nothing in either Vonoyman's

02:17:45.040 --> 02:17:47.677
background or Oppenheimer's background
or any of these people's background that

02:17:47.760 --> 02:17:49.597
qualifies them as moral authorities.

02:17:49.680 --> 02:17:54.157
>> Yeah. Well, this actually brings up the
point of in AI, who are the good people

02:17:54.240 --> 02:17:58.557
to to reason about the morality,
the ethics, the outside of these risks,

02:17:58.640 --> 02:18:02.237
outside like [snorts] the more
complicated stuff that you you agree on

02:18:02.320 --> 02:18:06.157
is, you know, this will go into
the hands of bad guys and all the kinds of

02:18:06.240 --> 02:18:08.157
ways they'll do is is interesting and

02:18:08.240 --> 02:18:13.437
dangerous. um is dangerous
in interesting unpredictable ways and who

02:18:13.520 --> 02:18:17.117
is the right person who are the right
kinds of people to make decisions how to

02:18:17.200 --> 02:18:19.597
respond to it is it tech people.

02:18:19.680 --> 02:18:22.797
>> So the history of these fields this
is what he talks about in the book the

02:18:22.880 --> 02:18:27.277
history of these fields is that the
the competence and capability and

02:18:27.360 --> 02:18:30.637
intelligence and training and
accomplishments of senior scientists and

02:18:30.720 --> 02:18:34.957
technologists working on a technology
and then being able to then make moral

02:18:35.040 --> 02:18:36.317
judgments in the use of the technology.

02:18:36.400 --> 02:18:39.997
That track record is terrible.
that track that track record is like

02:18:40.080 --> 02:18:43.837
catastrophically bad. Um >>
the people just the people that develop

02:18:43.920 --> 02:18:47.917
that technology are usually not
going to be the right people.

02:18:48.000 --> 02:18:50.477
>> Well, why would they So, the claim is
of course they're the knowledgeable ones,

02:18:50.560 --> 02:18:53.917
but the the problem is they've spent
their entire life in a lab, right?

02:18:54.000 --> 02:18:58.397
They're not theologians. Well, so what
you find what you find when you read

02:18:58.480 --> 02:19:00.477
when you read this and when you look
at these histories, what you find is they

02:19:00.560 --> 02:19:01.997
generally are very thinly informed on

02:19:02.080 --> 02:19:05.117
history. Yeah. on sociology, on on on um

02:19:05.200 --> 02:19:08.637
theology, on morality, on ethics. They

02:19:08.720 --> 02:19:12.237
they tend to manufacture their own
worldviews from scratch. They tend to be

02:19:12.320 --> 02:19:17.117
very sort of thin. um

02:19:17.200 --> 02:19:19.917
they're not remotely the arguments
that you would be having if you got like a

02:19:20.000 --> 02:19:23.597
group of highly qualified theologians
or philosophers or you know um

02:19:23.680 --> 02:19:25.837
>> well let me uh sort
of uh as the devil's

02:19:25.920 --> 02:19:29.597
advocate takes a sip of whiskey say that

02:19:29.680 --> 02:19:33.917
I I agree with uh with that but also it

02:19:34.000 --> 02:19:38.317
seems like the people who are doing kind
of the ethics departments and these text

02:19:38.400 --> 02:19:40.637
tech companies go sometimes the other

02:19:40.720 --> 02:19:41.117
way.

02:19:41.200 --> 02:19:45.757
>> Yes. uh they're not nuanced on the on

02:19:45.840 --> 02:19:47.917
history or theology or this kind of

02:19:48.000 --> 02:19:50.477
stuff. They it almost becomes kind of

02:19:50.560 --> 02:19:54.477
outraged activism towards um directions

02:19:54.560 --> 02:19:56.717
that don't seem to be Yeah. grounded in

02:19:56.800 --> 02:20:00.077
history and uh humility and nuance. It's

02:20:00.160 --> 02:20:04.637
again drenched with arrogance.
So >> I'm not sure which is worse.

02:20:04.720 --> 02:20:07.437
>> Well, no, they're both bad. Yeah.
So definitely not them either. Um

02:20:07.520 --> 02:20:10.317
>> so but I guess >>
but look, this is a hard

02:20:10.400 --> 02:20:13.037
>> Yeah, it's a hard problem. This is
our problem and this goes back to where we

02:20:13.120 --> 02:20:17.037
started which is okay who has the truth
and it's like well um you know like how

02:20:17.120 --> 02:20:19.917
do societies arrive at like truth
and how do we figure these things out and

02:20:20.000 --> 02:20:24.237
like our elected leaders play some role
in it you know we all play some role in

02:20:24.320 --> 02:20:28.317
it um there have to be some set
of public intellectuals at some point that

02:20:28.400 --> 02:20:31.197
bring you know rationality
and judgment humility to it

02:20:31.280 --> 02:20:34.477
>> those people are few and far between
we should probably prize them very highly

02:20:34.560 --> 02:20:36.717
>> yeah c celebrate humility in our public

02:20:36.800 --> 02:20:40.077
leaders uh so getting to risk number two

02:20:40.160 --> 02:20:44.317
will AI I ruin our society.
Short version as you write, if the murder

02:20:44.400 --> 02:20:48.797
robots don't get us, the hate
speech and misinformation will.

02:20:48.880 --> 02:20:52.477
>> And uh the action you recommend
in short, don't let the thought police

02:20:52.560 --> 02:20:55.357
oppress AI.

02:20:55.440 --> 02:21:00.397
>> Well, what is uh
this risk of the effect

02:21:00.480 --> 02:21:03.597
of [snorts] misinformation of society

02:21:03.680 --> 02:21:06.557
that's going to be catalyzed by AI?

02:21:06.640 --> 02:21:08.317
>> Yeah. So,
this is the social media. This

02:21:08.400 --> 02:21:11.437
is what you just alluded is the activism
kind of thing that's popped up in these

02:21:11.520 --> 02:21:14.637
companies in in the industry. And it's
basically from my perspective, it's

02:21:14.720 --> 02:21:17.677
basically part two of the war that
played out over social media over the

02:21:17.760 --> 02:21:21.437
last 10 years. Um, because you probably
remember social media 10 years ago was

02:21:21.520 --> 02:21:25.117
basically who even wants this? Who wants
who wants a photo of what your cat had

02:21:25.200 --> 02:21:27.837
for breakfast? Like this stuff is like
silly and trivial and why can't these

02:21:27.920 --> 02:21:31.117
nerds like figure out how to invent
something like useful and powerful? And

02:21:31.200 --> 02:21:34.797
then you know certain things happened in
the political system and then it sort of

02:21:34.880 --> 02:21:37.837
the polarity on that discussion switched
all the way to social media is like the

02:21:37.920 --> 02:21:41.277
worst most corrosive most terrible most
awful technology ever invented and it

02:21:41.360 --> 02:21:44.957
leads to you know terrible the wrong
you know politicians and policies and

02:21:45.040 --> 02:21:48.397
politics and like and all this stuff
and and that that all got catalyzed into

02:21:48.480 --> 02:21:52.317
this very big kind of angry movement
both inside and outside the companies to

02:21:52.400 --> 02:21:55.117
kind of bring social media to to to heal.
And that got focused in

02:21:55.200 --> 02:21:57.437
particularly on two topics
so-called hate speech and so-called

02:21:57.520 --> 02:22:00.717
misinformation. Um, and and that's been
this saga playing out for the last for

02:22:00.800 --> 02:22:03.517
the last decade. And I don't even really
want to even argue the pros and cons of

02:22:03.600 --> 02:22:06.477
the sides just to observe that that's
been like a huge fight and has had, you

02:22:06.560 --> 02:22:07.757
know, big consequences to how these

02:22:07.840 --> 02:22:11.197
companies operate. Um, basically that

02:22:11.280 --> 02:22:15.277
same those same sets of theories, that
same activist approach, that same energy

02:22:15.360 --> 02:22:17.517
is being transplanted straight to AI.

02:22:17.600 --> 02:22:20.477
And you see that already happening.
It's why, you know, Chad GPT will answer

02:22:20.560 --> 02:22:23.757
let's say certain questions and not
others. Um it's why it gives you the can

02:22:23.840 --> 02:22:27.277
speech about you know whenever it starts
with as a large language model I cannot

02:22:27.360 --> 02:22:29.677
you know basically means that somebody
has reached in there and told that it

02:22:29.760 --> 02:22:33.357
can't talk about certain topics.
Um >> do you think some of that is good?

02:22:33.440 --> 02:22:35.277
>> So it's a it's an interesting question.

02:22:35.360 --> 02:22:39.917
Um so a couple couple observations. Um
so so one is um the people who find this

02:22:40.000 --> 02:22:43.487
the most frustrating are the people
who are worried about the murder robots.

02:22:43.570 --> 02:22:46.877
[laughter] >> Right.
So so and in fact the the ex

02:22:46.960 --> 02:22:50.797
so-called X-risk people right they
started with the term AI safety. the

02:22:50.880 --> 02:22:53.917
term became AI alignment. When the term
became AI alignment is when this switch

02:22:54.000 --> 02:22:55.997
happened from we're worried it's going
to kill us all to we're worried about

02:22:56.080 --> 02:22:56.587
hate speech and misinformation.

02:22:56.670 --> 02:22:59.837
[clears throat] >> Sure.
>> The AIX risk people have now renamed

02:22:59.920 --> 02:23:03.277
their thing uh AI not
kill everyone ism [snorts]

02:23:03.360 --> 02:23:06.957
which I have to admit is a catchy term
and they are very frustrated by the fact

02:23:07.040 --> 02:23:09.917
that the hate the sort of activist
driven hate speech misinformation kind

02:23:10.000 --> 02:23:12.637
of thing is taking over which is what's
happened. It's taken over the AI ethics

02:23:12.720 --> 02:23:16.317
field has been taken over by the hate
speech misinformation people. Um, you

02:23:16.400 --> 02:23:19.677
know, look, would I like to live in a
world in which like everybody was nice

02:23:19.760 --> 02:23:22.477
to each other all the time and nobody
ever said anything mean and nobody ever

02:23:22.560 --> 02:23:25.597
used a bad word and everything
was always accurate and honest? Like that

02:23:25.680 --> 02:23:28.317
sounds great. Do I want to live in a
world where there's like a centralized

02:23:28.400 --> 02:23:32.077
thought police working through the tech
companies to enforce the view of a small

02:23:32.160 --> 02:23:34.557
set of elites that they're going
to determine what the rest of us think and

02:23:34.640 --> 02:23:38.157
feel like? Absolutely not. >> There
could be a middle ground somewhere

02:23:38.240 --> 02:23:39.757
like Wikipedia type of moderation.

02:23:39.840 --> 02:23:42.397
there's moderation on Wikipedia

02:23:42.480 --> 02:23:45.357
that it's somehow crowdsourced where you

02:23:45.440 --> 02:23:48.317
don't have centralized elites. Uh but

02:23:48.400 --> 02:23:50.317
it's also not completely just a

02:23:50.400 --> 02:23:53.357
free-for-all because uh the if you have

02:23:53.440 --> 02:23:55.597
the entirety of human knowledge at your

02:23:55.680 --> 02:23:58.477
fingertips. You can do a lot of harm.

02:23:58.560 --> 02:24:00.797
Like if if you have a good assistant

02:24:00.880 --> 02:24:03.197
that's completely uncensored, they can

02:24:03.280 --> 02:24:06.477
help you build a bomb. they can help you

02:24:06.560 --> 02:24:11.277
um mess with people's physical well-being,

02:24:11.360 --> 02:24:13.197
right? If they because that information

02:24:13.280 --> 02:24:14.957
is out there on the internet. And so

02:24:15.040 --> 02:24:19.197
there presumably there's it would be you

02:24:19.280 --> 02:24:22.637
could see the positives in um censoring

02:24:22.720 --> 02:24:27.437
some aspects of an AI model when it's
helping you commit literal violence.

02:24:27.520 --> 02:24:30.157
>> Yeah. And there's a section later
section of the essay where I talk about

02:24:30.240 --> 02:24:31.517
bad people doing bad things. Yes.

02:24:31.600 --> 02:24:33.437
>> Right. which which
and there's a there's

02:24:33.520 --> 02:24:34.957
a set of things that we
should discuss there.

02:24:35.040 --> 02:24:38.477
>> Yeah. Um, what happens in practice
is these line, as you alluded to this

02:24:38.560 --> 02:24:41.277
already, these lines are not easy to draw.
And what what I've observed in the

02:24:41.360 --> 02:24:44.637
social media version of this is the way
I describe it as the slippery slope is

02:24:44.720 --> 02:24:46.557
not a fallacy. It's an inevitability.

02:24:46.640 --> 02:24:49.197
The minute you have this kind
of activist personality that gets in a

02:24:49.280 --> 02:24:52.077
position [clears throat] to make these
decisions, they they take it straight to

02:24:52.160 --> 02:24:55.997
infinity. Like they it goes into
the crazy zone like almost immediately and

02:24:56.080 --> 02:24:59.197
never comes back because people become
drunk with power. Um, right. And they

02:24:59.280 --> 02:25:01.757
they look if you're in the position
to determine what the entire world thinks

02:25:01.840 --> 02:25:05.917
and feels and reads and says like you're
going to take it. And you know Elon has

02:25:06.000 --> 02:25:08.477
you know ventilated this with the
Twitter files over the last you know 3

02:25:08.560 --> 02:25:11.437
months and it's just like crystal
clear like how bad it got there. Now

02:25:11.520 --> 02:25:16.477
>> reason for optimism is what Elon
is doing with community notes. Um um so

02:25:16.560 --> 02:25:19.997
community notes is actually a very
interesting thing. U so what Elon is

02:25:20.080 --> 02:25:22.717
trying to do with community notes um
is he's trying to have it where there's

02:25:22.800 --> 02:25:26.637
only a community note when people who
have previously disagreed on many topics

02:25:26.720 --> 02:25:31.117
agree on this one. Yes, that's that's
what that's what I'm trying to get at is

02:25:31.200 --> 02:25:32.717
like there's there could be Wikipedia

02:25:32.800 --> 02:25:34.637
like models or community notes type of

02:25:34.720 --> 02:25:38.317
models where allows you to essentially

02:25:38.400 --> 02:25:43.037
either provide context or censor in a
way that does not resist the slippery

02:25:43.120 --> 02:25:45.517
slope nature. >> Now there's
another there's an entirely

02:25:45.600 --> 02:25:48.717
different approach here which
is basically um we have AIs that are

02:25:48.800 --> 02:25:51.677
producing content we could also
have AIs that are consuming content.

02:25:51.760 --> 02:25:53.117
>> Yeah. >> Right.
And so one of the things that

02:25:53.200 --> 02:25:57.197
your assistant could do for you is help
you consume all the content. Right. and

02:25:57.280 --> 02:26:00.317
basically tell you when you're getting
played. So, for example, I'm going to

02:26:00.400 --> 02:26:03.837
want the AI that my kid uses, right,
to be very, you know, child safe and I'm

02:26:03.920 --> 02:26:06.557
going to want it to filter for him all
kinds of inappropriate stuff that he

02:26:06.640 --> 02:26:08.077
shouldn't be saying just cuz he's a kid.

02:26:08.160 --> 02:26:09.597
>> Yeah. >> Right.
And you see what I'm saying is

02:26:09.680 --> 02:26:11.997
you can implement that. You could you
architecturally you could say you can

02:26:12.080 --> 02:26:13.597
solve this on the client side. Right.

02:26:13.680 --> 02:26:16.477
Solving on the server side gives you
an opportunity to dictate for the entire

02:26:16.560 --> 02:26:20.077
world, which I think is where you you
take the slippery slope to hell. Um

02:26:20.160 --> 02:26:22.477
there's another architectural approach
which is to solve this on the client

02:26:22.560 --> 02:26:26.877
side, which is certainly what I would
endorse. It's AI risk number five. Will

02:26:26.960 --> 02:26:30.877
AI lead uh to bad people doing bad things?
And I can just imagine language

02:26:30.960 --> 02:26:32.557
models used to do so many bad things.

02:26:32.640 --> 02:26:36.077
But the hope is there that you can have

02:26:36.160 --> 02:26:39.757
uh large language models used to then
defend against it by more people, by

02:26:39.840 --> 02:26:42.877
smarter people, by um more effective

02:26:42.960 --> 02:26:46.797
people, skilled people, all that kind
of stuff. Three three-part argument on bad

02:26:46.880 --> 02:26:50.077
people doing bad things. Um so um uh
so number one right you can use the

02:26:50.160 --> 02:26:53.277
technology defensively and there's a we
should be using AI to build like broadsp

02:26:53.360 --> 02:26:56.637
spectrum vaccines and antibiotics
for like bioweapons and we should be using

02:26:56.720 --> 02:26:59.357
AI to like hunt terrorists and catch
criminals and like we should be doing

02:26:59.440 --> 02:27:02.077
like all kinds of stuff like that and
in fact we should be doing those things

02:27:02.160 --> 02:27:05.037
even just to like go get like you know
basically go eliminate risk from like

02:27:05.120 --> 02:27:07.757
regular pathogens that aren't like
constructed by an AI. So there's there's

02:27:07.840 --> 02:27:11.997
there's the whole um uh there's a whole
defensive set of things. Um second is we

02:27:12.080 --> 02:27:15.597
have many laws on the books about
the actual bad things right so it is

02:27:15.680 --> 02:27:19.597
actually illegal to be a you know
to commit crimes to commit terrorist acts

02:27:19.680 --> 02:27:23.437
to you know build pathogens with the
intent to deploy them to kill people and

02:27:23.520 --> 02:27:26.877
so we have those we we don't we
actually don't need new laws for the vast

02:27:26.960 --> 02:27:30.077
majority of these scenarios we actually
already have the laws on the book on the

02:27:30.160 --> 02:27:33.917
books. The third argument is the minute
and this is sort of the foundational one

02:27:34.000 --> 02:27:36.797
that gets really tough but the minute
you get into this thing which which you

02:27:36.880 --> 02:27:39.517
were kind of getting into which is like
okay but like don't you need censorship

02:27:39.600 --> 02:27:42.237
sometimes right and don't you need
restrictions sometimes it's like okay

02:27:42.320 --> 02:27:44.957
what is the cost of that um and in

02:27:45.040 --> 02:27:46.717
particular in the world of open source

02:27:46.800 --> 02:27:49.757
right um and so um is open source AI

02:27:49.840 --> 02:27:52.557
going to be allowed or not um if open

02:27:52.640 --> 02:27:57.437
source AI is not allowed um then what
is the regime that's going to be necessary

02:27:57.520 --> 02:28:00.637
legally and technically
to prevent it from developing

02:28:00.720 --> 02:28:03.837
Right. And here again is where you get
into and people have proposed that these

02:28:03.920 --> 02:28:06.477
kinds of things. You get into, I would
say, pretty extreme territory pretty

02:28:06.560 --> 02:28:10.957
fast. Do we have a monitor agent
on every CPU and GPU that reports back to

02:28:11.040 --> 02:28:15.037
the government what we're doing with our
computers? Are we seizing GPU clusters

02:28:15.120 --> 02:28:18.237
that get beyond a certain size? Like,
and then by the way, how are we doing

02:28:18.320 --> 02:28:22.557
all that globally, right? And like if
China is developing an LLM beyond the

02:28:22.640 --> 02:28:25.757
scale that we think is allowable,
are we going to invade?

02:28:25.840 --> 02:28:28.957
>> Right? Right? And you have figures
on the AIX risk side who are advocating,

02:28:29.040 --> 02:28:31.517
you know, potentially up to nuclear
strikes to prevent, you know, this kind

02:28:31.600 --> 02:28:35.597
of thing. And so here you get into
this thing and again, you could maybe say

02:28:35.680 --> 02:28:38.237
this is, you know, you could even say
this is what good, bad or indifferent or

02:28:38.320 --> 02:28:41.997
whatever. But like here's the the
comparison of nukes. The comparison of

02:28:42.080 --> 02:28:45.437
nukes is very dangerous because one
is just nukes were just just a bomb.

02:28:45.520 --> 02:28:47.757
Although we can come back to nuclear
power, but the other thing was like with

02:28:47.840 --> 02:28:50.637
nukes you could control plutonium,
right? You could track plutonium and it

02:28:50.720 --> 02:28:54.957
was like hard to come by. [snorts] AI is
just math and code, right? It's and it's

02:28:55.040 --> 02:28:57.357
in like math textbooks and it's like
there are YouTube videos that teach you

02:28:57.440 --> 02:28:59.677
how to build it and like there's
open it's already open source you know

02:28:59.760 --> 02:29:02.877
there's a 40 billion parameter model
running around already called Falcon

02:29:02.960 --> 02:29:07.997
online that anybody can download. Um and
so okay you you walk down the logic path

02:29:08.080 --> 02:29:11.437
that says we need to have guardrails
on this and you find yourself in a

02:29:11.520 --> 02:29:13.517
authoritarian totalitarian regime of

02:29:13.600 --> 02:29:19.197
thought control and machine control that
would be so brutal that you would have

02:29:19.280 --> 02:29:22.477
destroyed the society that you're trying
to protect. And so I I I just don't see

02:29:22.560 --> 02:29:26.237
how that actually works.
>> So you have to understand my brain's

02:29:26.320 --> 02:29:30.717
going full uh full steam ahead here
because I agree with uh basically

02:29:30.800 --> 02:29:33.917
everything you're saying, but I'm trying
to play devil's advocate here there

02:29:34.000 --> 02:29:38.077
because okay, you highlighted [snorts]
the fact that there is a slippery slope

02:29:38.160 --> 02:29:42.077
to human nature. The moment you
censor something, you start to censor

02:29:42.160 --> 02:29:43.837
everything.

02:29:43.920 --> 02:29:48.317
uh that alignment starts out sounding

02:29:48.400 --> 02:29:51.597
nice, but then you start to align to uh

02:29:51.680 --> 02:29:54.797
the beliefs of some select group of

02:29:54.880 --> 02:29:57.037
people and then it's just your beliefs.

02:29:57.120 --> 02:30:00.637
This the the number the number of people
you're aligning to smaller and smaller

02:30:00.720 --> 02:30:05.037
as that group becomes more and more
powerful. Okay, but that just speaks to

02:30:05.120 --> 02:30:10.157
the people that censor are
usually the and uh the get richer.

02:30:10.240 --> 02:30:11.757
I wonder if it's possible to do without

02:30:11.840 --> 02:30:14.957
that for AI. One way to ask this

02:30:15.040 --> 02:30:18.317
question is do you think the base models

02:30:18.400 --> 02:30:20.717
the the base the baseline foundation

02:30:20.800 --> 02:30:23.677
models should be open sourced like uh

02:30:23.760 --> 02:30:28.637
where the Mark Zuckerberg is saying they
want to do? So I look I I mean I think

02:30:28.720 --> 02:30:32.317
it's totally appropriate that companies
that are in the business of producing a

02:30:32.400 --> 02:30:35.917
product or service should be able
to have a wide range of policies that they

02:30:36.000 --> 02:30:40.637
put right and I just again I want a
heavily censored model for my 8-year-old.

02:30:40.720 --> 02:30:43.197
>> Like I actually want that like like
I would pay more money for the one that's

02:30:43.280 --> 02:30:46.477
more heavily censored than the one
that's not right. Um and so like there

02:30:46.560 --> 02:30:49.917
are certainly scenarios where companies
will make that decision. Look an

02:30:50.000 --> 02:30:53.757
interesting thing you brought up though
or is is this really a speech issue? Um,

02:30:53.840 --> 02:30:56.397
one of the things that the big tech
companies are dealing with is that

02:30:56.480 --> 02:30:59.037
content generated uh from an LLM is not

02:30:59.120 --> 02:31:03.677
covered under section 230 uh which
is the law that u protects internet

02:31:03.760 --> 02:31:05.277
platform companies from being sued for

02:31:05.360 --> 02:31:08.397
the user generated content. Um, and so

02:31:08.480 --> 02:31:12.557
it it's actually Yes. And so there's
actually there's actually a question I

02:31:12.640 --> 02:31:15.357
think there's still a question which
is can big can big American companies

02:31:15.440 --> 02:31:19.277
actually feel generative AI at all
or is the liability actually going to just

02:31:19.360 --> 02:31:22.797
ultimately convince them that they can't
do it because the minute the thing says

02:31:22.880 --> 02:31:25.597
something bad and it doesn't even need
to be hate speech. It could just be like

02:31:25.680 --> 02:31:30.077
an inacc It could hallucinate a product,
you know, detail on a vacuum cleaner,

02:31:30.160 --> 02:31:32.877
you know, and all of a sudden
the vacuum cleaner company sues for

02:31:32.960 --> 02:31:36.157
misrepresentation and there's any
symmetry there, right? Cuz the the the

02:31:36.240 --> 02:31:38.957
LM is going to be producing billions
of answers to questions and it only needs

02:31:39.040 --> 02:31:41.677
to get a few wrong to >> So loss
has to get updated really quick here.

02:31:41.760 --> 02:31:44.957
>> Yeah. And nobody knows what to do with
that, right? Um so so anyway, like there

02:31:45.040 --> 02:31:47.757
there there are big there are big
questions around how companies operate

02:31:47.840 --> 02:31:51.197
at all. So we we talk about those,
but then there's this other question of

02:31:51.280 --> 02:31:53.997
like, okay, the open source, so what
about open source? And and my answer to

02:31:54.080 --> 02:31:58.077
your question is kind of like obviously
yes the models have there has to be full

02:31:58.160 --> 02:32:01.917
open source here because to live in a
world in which that open source is not

02:32:02.000 --> 02:32:05.037
allowed is a world of draconian speech

02:32:05.120 --> 02:32:09.677
control human control machine control.
I mean you know black helicopters with

02:32:09.760 --> 02:32:14.157
jack booted thugs coming out repelling
down and seizing your GPU like territory.

02:32:14.240 --> 02:32:16.637
>> Well >> no no I'm 100% serious.

02:32:16.720 --> 02:32:18.717
>> I that's you're saying
slippery slope always leaves there.

02:32:18.800 --> 02:32:21.917
>> No no no no no. That's what's required
to enforce it. Like how will you enforce

02:32:22.000 --> 02:32:25.757
a ban on open source? >> No,
you could add friction to it like

02:32:25.840 --> 02:32:28.877
harder to get the models cuz people
will always be able to get the models but

02:32:28.960 --> 02:32:30.557
it'll be more in the shadows, right?

02:32:30.640 --> 02:32:34.237
>> The leading open source model right now
is from the UAE. Like the next time they

02:32:34.320 --> 02:32:37.677
do that, what do we do? >> Yeah.

02:32:37.760 --> 02:32:42.797
>> Like, >> oh, I see you're like u >>
the 14-year-old in Indonesia comes out

02:32:42.880 --> 02:32:45.437
with a breakthrough. You know, we talked
about most great software comes from a

02:32:45.520 --> 02:32:47.997
small number of people. some kid comes
out with some big new breakthrough in

02:32:48.080 --> 02:32:51.197
quantization or something and has some
huge breakthrough and like what we're

02:32:51.280 --> 02:32:54.397
going to what are we going to like
invade Indonesia and arrest him.

02:32:54.480 --> 02:32:55.917
>> It seems like
in terms of size of models

02:32:56.000 --> 02:32:59.917
and effectiveness of models the big tech
companies will probably lead the way for

02:33:00.000 --> 02:33:01.757
quite a few years and and the question

02:33:01.840 --> 02:33:04.877
is of what policies they should use. The

02:33:04.960 --> 02:33:07.597
the kid the kid in Indonesia should not

02:33:07.680 --> 02:33:12.477
be regulated but should Google Meta

02:33:12.560 --> 02:33:17.837
uh Microsoft OpenAI be regulated? Well,
so but this goes okay. So when does it

02:33:17.920 --> 02:33:22.157
become dangerous? >> Yeah. >> Right.
Is is the danger that it's quote

02:33:22.240 --> 02:33:26.397
as powerful as the current leading
commercial model or is it that it is it

02:33:26.480 --> 02:33:28.397
is just at some other arbitrary threshold?

02:33:28.480 --> 02:33:30.477
>> Yeah. >> And then
by the way like look how do we

02:33:30.560 --> 02:33:33.677
know like what we know today is that you
need like a lot of money to like train

02:33:33.760 --> 02:33:36.077
these things. But there are advances
being made every week on training

02:33:36.160 --> 02:33:39.117
efficiency and you know data all kinds
of synthetic you know look I don't even

02:33:39.200 --> 02:33:41.517
like the synthetic data thing we're
talking about. Maybe some kid figures

02:33:41.600 --> 02:33:42.877
out a way to autogenerate synthetic data.

02:33:42.960 --> 02:33:43.997
>> That's going to change everything.

02:33:44.080 --> 02:33:47.677
>> Yeah. Exactly. And so like sitting here
today like the the the breakthrough just

02:33:47.760 --> 02:33:50.477
happened, right? You made this point
like the breakthrough just happened.

02:33:50.560 --> 02:33:54.077
>> So we don't know what the shape of
this technology is going to be. I mean the

02:33:54.160 --> 02:33:58.477
the big shock the the big shock here
is that you know whatever number of

02:33:58.560 --> 02:34:01.597
billions of parameters basically
represents at least a very big

02:34:01.680 --> 02:34:05.597
percentage of human thought.
>> Like who would have imagined that? And

02:34:05.680 --> 02:34:06.957
then there's already work underway.

02:34:07.040 --> 02:34:10.397
There was just this paper that just came
out that basically takes a GPT3 scale

02:34:10.480 --> 02:34:13.917
model and compresses it down
to run on a single 32 core CPU.

02:34:14.000 --> 02:34:15.997
>> Like who would have predicted that?

02:34:16.080 --> 02:34:18.477
>> Yeah. >> Um you know
some of these models now you

02:34:18.560 --> 02:34:21.677
can run on Raspberry Pies. Like today
they're very slow but like you know

02:34:21.760 --> 02:34:25.437
maybe they'll be a you know
real perform you know like

02:34:25.520 --> 02:34:29.277
>> it's math and here we're back in here
we're back dude. It's math and code.

02:34:29.360 --> 02:34:31.997
It's math and code.
It's math code and data. It's bits.

02:34:32.080 --> 02:34:36.637
>> Mark's just like walked away [laughter]
at this point. He's just screw it. I

02:34:36.720 --> 02:34:40.317
don't know what to do with this. You
guys created this whole internet thing.

02:34:40.400 --> 02:34:44.397
Yeah. Yeah. I'm a huge
believer in open source here.

02:34:44.480 --> 02:34:47.197
>> So my argument is we're going to have
to see here's my argument is my argument.

02:34:47.280 --> 02:34:49.837
My full argument is AI is going to be
like air. It's going to be everywhere.

02:34:49.920 --> 02:34:52.157
Like it's this is just going to be in
text. It already is. It's going to be in

02:34:52.240 --> 02:34:54.157
textbooks and kids are going to grow
up knowing how to do this and it's just

02:34:54.240 --> 02:34:56.477
going to be a thing. It's going to be
in the air and you can't like pull this

02:34:56.560 --> 02:34:59.277
back anymore. You can pull back air.
And so you just have to figure out how to

02:34:59.360 --> 02:35:02.077
live in this world, right? Right.
And then that that and then that's where I

02:35:02.160 --> 02:35:04.797
think like all this hand ringing about
AI risk is basically a complete waste of

02:35:04.880 --> 02:35:08.157
time because the the the effort should
go into okay what are what what is what

02:35:08.240 --> 02:35:11.117
what is the defensive approach >>
and so if you're worried about you know

02:35:11.200 --> 02:35:13.997
AI generated pathogens the right thing
to do is to have a permanent project

02:35:14.080 --> 02:35:17.517
warp speed right funded lavishly let's
let's do a Manhattan let's talk about

02:35:17.600 --> 02:35:20.957
Manhattan project let's do a Manhattan
project for biological defense right and

02:35:21.040 --> 02:35:23.917
let's build AIS and let's have like
broadspectctrum vaccines where like

02:35:24.000 --> 02:35:26.077
we're insulated from every pathogen

02:35:26.160 --> 02:35:28.957
>> right and what the interesting thing is

02:35:29.040 --> 02:35:31.677
because it's software where a kid in his

02:35:31.760 --> 02:35:34.957
basement teenager could build like a

02:35:35.040 --> 02:35:36.237
system that defends against like the

02:35:36.320 --> 02:35:39.357
worst the the worst I mean and to me

02:35:39.440 --> 02:35:42.477
defense is super exciting.

02:35:42.560 --> 02:35:47.277
>> It's to like I if you believe in the
good of human nature that most people

02:35:47.360 --> 02:35:51.357
want to do good to be the savior
of humanity is really exciting.

02:35:51.440 --> 02:35:54.637
>> Yes. [laughter] >> Not okay
that's a dramatic statement but

02:35:54.720 --> 02:35:57.277
like to help people. To help people

02:35:57.360 --> 02:35:59.357
>> Yeah. Okay. What about just to jump

02:35:59.440 --> 02:36:04.557
around, what about the risk of will
AI lead to crippling inequality?

02:36:04.640 --> 02:36:08.397
You know, because we're kind of saying
everybody's life will become better. Is

02:36:08.480 --> 02:36:10.477
it possible that the
the rich get richer here?

02:36:10.560 --> 02:36:13.757
>> Yeah. So, this go this actually
ironically goes back to Marxism. So, um

02:36:13.840 --> 02:36:16.477
because this was the C. So, the core
claim of Marxism, right, basically was

02:36:16.560 --> 02:36:19.037
that the owner the owners of capital
would basically own the means of

02:36:19.120 --> 02:36:22.077
production and then over time they would
basically accumulate all the wealth. The

02:36:22.160 --> 02:36:25.037
workers would be paying in, you know,
and and getting nothing in return

02:36:25.120 --> 02:36:26.317
because they wouldn't be needed anymore.

02:36:26.400 --> 02:36:28.797
Right? Marx was very worried about me
what he called mechanization or what

02:36:28.880 --> 02:36:32.317
later became known as automation um
and that you know the workers would be

02:36:32.400 --> 02:36:35.437
emiserated and the the capitalist would
end up with with with all and so this

02:36:35.520 --> 02:36:38.397
was one of the core core core principles
of Marxism of course it turned out to be

02:36:38.480 --> 02:36:41.837
wrong about every previous wave
of technology um the reason it it turned

02:36:41.920 --> 02:36:45.197
out to be wrong about every previous
wave of technology is that the way that

02:36:45.280 --> 02:36:48.477
the self-interested owner of the
machines makes the most money is by

02:36:48.560 --> 02:36:52.957
providing the production capability in
the form of products and services to the

02:36:53.040 --> 02:36:57.037
most people the most customers
as possible right the the largest, this is

02:36:57.120 --> 02:36:59.597
one of those funny things where every
CEO knows this intuitively and yet it's

02:36:59.680 --> 02:37:01.277
like hard to explain from the outside.

02:37:01.360 --> 02:37:04.077
The the way you make the most money
in any business is by selling to the

02:37:04.160 --> 02:37:05.837
largest market you can possibly get to.

02:37:05.920 --> 02:37:09.437
The largest market you can possibly get
to is everybody on the planet. And so

02:37:09.520 --> 02:37:12.957
every large company does is everything
that it can to drive down prices to be

02:37:13.040 --> 02:37:15.917
able to get volumes up to be able to get
to everybody on the planet. And that

02:37:16.000 --> 02:37:18.317
happened with everything from
electricity. It happened with

02:37:18.400 --> 02:37:21.277
telephones. It happened with radio.
It happened with automobiles. It happened

02:37:21.360 --> 02:37:23.997
with smartphones. It happened with

02:37:24.080 --> 02:37:26.637
PCs. Um it happened with the internet.

02:37:26.720 --> 02:37:30.797
Um it happened with mobile broadband. Um
it's happened by the way with Coca-Cola.

02:37:30.880 --> 02:37:34.477
It's happened with like every, you know,
basically every industrially produced,

02:37:34.560 --> 02:37:36.877
you know, good or service. People want
you want to drive it to the largest

02:37:36.960 --> 02:37:40.477
possible market. And then as proof
of that, it's already happened, right?

02:37:40.560 --> 02:37:43.357
which is the early adopters of like Jad

02:37:43.440 --> 02:37:47.597
GPT and Bing are not like, you know,
Exxon and Boeing. They're, you know,

02:37:47.680 --> 02:37:51.677
your uncle and your nephew, right? It's
just like it's either freely available

02:37:51.760 --> 02:37:54.557
online or it's available for 20 bucks
a month or something. But the, you know,

02:37:54.640 --> 02:37:58.157
these things went this this this
technology went mass market immediately.

02:37:58.240 --> 02:38:01.597
Um, and so look, the the the owners
of the means of production, whoever does

02:38:01.680 --> 02:38:03.677
this, you mentioned these trillion
dollar questions, there are people who

02:38:03.760 --> 02:38:06.637
are going to get really rich doing this,
producing these things, but they're

02:38:06.720 --> 02:38:08.957
going to get really rich by taking
this technology to the broadest possible

02:38:09.040 --> 02:38:13.997
market. So yes, they'll get rich, but
they'll get rich having a huge positive

02:38:14.080 --> 02:38:17.117
impact on >> Yeah. making the making
the technology available to everybody.

02:38:17.200 --> 02:38:18.717
>> Yeah. >> Right.
And again, smartphones, same

02:38:18.800 --> 02:38:22.717
thing. So, and there's this amazing
kind of twist in um in business history,

02:38:22.800 --> 02:38:26.797
which is you cannot spend $10,000
on a smartphone, right? You can't spend

02:38:26.880 --> 02:38:29.517
$100,000. You can't spend a like I would
buy the million dollar smartphone. Like

02:38:29.600 --> 02:38:32.157
I'm signed up for it. Like if it's like
suppose a million dollar smartphone was

02:38:32.240 --> 02:38:35.357
like much better than the $1,000
smartphone. Like I'm there to buy it. It

02:38:35.440 --> 02:38:37.197
doesn't exist. Why doesn't it exist?

02:38:37.280 --> 02:38:40.397
Apple makes so much more money driving
the price further down from $1,000 than

02:38:40.480 --> 02:38:43.517
they would trying to harvest, right? And
so it's it's just this repeating pattern

02:38:43.600 --> 02:38:46.637
you see over and over again. Um where
the and and and what's what's great

02:38:46.720 --> 02:38:49.677
about it what's great about it is you
you do not need to rely on anybody's

02:38:49.760 --> 02:38:52.077
enlightened right generosity to do this.

02:38:52.160 --> 02:38:55.517
You just need to rely
on capitalist self-interest.

02:38:55.600 --> 02:38:58.397
>> Uh what about AI taking our jobs?

02:38:58.480 --> 02:39:00.157
>> Yeah, so very very similar thing here.

02:39:00.240 --> 02:39:03.037
Um there's sort of a there's a core
fallacy which again was was very common

02:39:03.120 --> 02:39:06.397
in Marxism which is what's called the
lump of labor fallacy. And this is sort

02:39:06.480 --> 02:39:09.517
of the fallacy that there is a only
a fixed amount of work to be done in the

02:39:09.600 --> 02:39:13.117
world and if the and it's all being done
today by people and then if machines do

02:39:13.200 --> 02:39:16.957
it there's no other work to be done
by people. Um and that's just a completely

02:39:17.040 --> 02:39:20.797
backwards view on how the economy
develops and grows. Um because what

02:39:20.880 --> 02:39:25.037
happens is not in fact that what happens
is the introduction of technology into

02:39:25.120 --> 02:39:29.197
production process causes prices to fall.
As prices fall, consumers have

02:39:29.280 --> 02:39:32.557
more spending power. As consumers have
more spending power, they create new

02:39:32.640 --> 02:39:37.037
demand. That new demand then causes
capital and labor to form into new

02:39:37.120 --> 02:39:40.797
enterprises to satisfy new wants
and needs and the result is more jobs and

02:39:40.880 --> 02:39:43.997
higher wages. >> So new wants
and needs the the worry is

02:39:44.080 --> 02:39:45.757
that the the creation of new wants and

02:39:45.840 --> 02:39:49.197
needs at a rapid rate will mean there's

02:39:49.280 --> 02:39:53.757
a lot of turnover in jobs.
So people will lose jobs. Just the actual

02:39:53.840 --> 02:39:57.277
experience of losing a job and having
to learn new things and new skills is

02:39:57.360 --> 02:39:59.917
painful for the individuals.
>> Well, two things. One is that new jobs

02:40:00.000 --> 02:40:03.757
are often much better. Um, so this
actually came up that there was this

02:40:03.840 --> 02:40:05.917
panic about a decade ago and all
the truck drivers are going to lose their

02:40:06.000 --> 02:40:08.557
jobs, right? And number one, that didn't
happen because we haven't figured out a

02:40:08.640 --> 02:40:12.557
way to actually finish that yet. But >>
but the other thing was like look, truck

02:40:12.640 --> 02:40:15.437
driver like I grew up in a town that
was basically consisted of a truck stop,

02:40:15.520 --> 02:40:18.477
right? And I like knew a lot of truck
drivers and like truck drivers live a

02:40:18.560 --> 02:40:22.557
decade shorter than everybody else. Like
they it's a it's a it's actually like a

02:40:22.640 --> 02:40:25.197
very dangerous like they get like
literally they have like high rates of

02:40:25.280 --> 02:40:27.917
skin cancer and on the left side
of their on the left side of their body

02:40:28.000 --> 02:40:29.597
from from being in the sun all the time.

02:40:29.680 --> 02:40:32.557
The vibration of being in the truck
is actually very damaging to your to your

02:40:32.640 --> 02:40:35.277
physiology.
>> And there's actually uh perhaps

02:40:35.360 --> 02:40:39.037
partially because of that reason
uh there's a shortage >> Yeah.

02:40:39.120 --> 02:40:42.077
>> of uh people who
want to be truck drivers.

02:40:42.160 --> 02:40:45.117
>> Yeah. Like it's not it's not like the
question always you want to ask somebody

02:40:45.200 --> 02:40:48.477
like that is do you want you know do you
want your kid to be doing this job? And

02:40:48.560 --> 02:40:51.437
like most of them will tell you no like
I want my kid to be sitting in a cubicle

02:40:51.520 --> 02:40:54.077
somewhere like where they don't have
this like where they don't die 10 years

02:40:54.160 --> 02:40:57.997
earlier. And so, so the new jobs, number
one, the new jobs are often better, but

02:40:58.080 --> 02:41:00.797
you don't get the new jobs until you
go through the change. And then to your

02:41:00.880 --> 02:41:03.837
point, the the training thing, you know,
is always the issue is can can people

02:41:03.920 --> 02:41:06.397
adapt? And again, here you need
to imagine living in a world in which

02:41:06.480 --> 02:41:10.237
everybody has the AI assistant capability,
right, to be able to pick up

02:41:10.320 --> 02:41:12.477
new skills much more quickly and be able
to have some, you know, be able to have

02:41:12.560 --> 02:41:14.077
a machine to work with
to augment their skills.

02:41:14.160 --> 02:41:17.117
>> It's still going to be painful,
but that's the process of life.

02:41:17.200 --> 02:41:19.357
>> It's painful for some people. I mean
there's no look there's no question it's

02:41:19.440 --> 02:41:22.877
painful for some people and there
you know yes it's not again I'm not a

02:41:22.960 --> 02:41:25.437
utopian on this and it's not like
it's it's positive for everybody in the

02:41:25.520 --> 02:41:29.757
moment but it has been overwhelmingly
positive for 300 years I mean look the

02:41:29.840 --> 02:41:32.797
concern here the concern the concern
this concern has played out for for

02:41:32.880 --> 02:41:36.157
literally centuries um and you know this
is the sort of lite you know the story

02:41:36.240 --> 02:41:41.037
of the leites um that you may remember
there was a panic in the 2000s around

02:41:41.120 --> 02:41:45.037
outsourcing was going to take all the jobs
there was a panic in the 2010s that

02:41:45.120 --> 02:41:47.037
robots were going to take all the jobs

02:41:47.120 --> 02:41:51.437
Um in 2019 before co we had more jobs at

02:41:51.520 --> 02:41:54.237
higher wages both in the country and
in the world than at any point in human

02:41:54.320 --> 02:41:57.597
history. >> And so
the overwhelming evidence is that

02:41:57.680 --> 02:42:02.317
the net gain here is like just like
wildly positive and most most people

02:42:02.400 --> 02:42:05.677
like overwhelmingly come out the other
side being huge beneficiaries of this.

02:42:05.760 --> 02:42:09.917
>> So you're right that the single
greatest risk this is the risk you're most

02:42:10.000 --> 02:42:14.397
convinced by. The single greatest
risk of AI is that China wins global AI

02:42:14.480 --> 02:42:18.237
dominance and we the United
States and the West do not.

02:42:18.320 --> 02:42:21.437
>> Can you elaborate? >> Yeah.
So, this is the other thing which

02:42:21.520 --> 02:42:25.037
is a lot of the sort of AI risk debates
today sort of assume that we're the only

02:42:25.120 --> 02:42:27.997
game in town, right? And so, we have
the ability to kind of sit in the United

02:42:28.080 --> 02:42:30.797
States and criticize ourselves and,
you know, have our government like, you

02:42:30.880 --> 02:42:33.197
know, beat up on our companies and we're
figure out a way to restrict what our

02:42:33.280 --> 02:42:35.517
companies can do and, you know, we're
going to, you know, we're going to ban

02:42:35.600 --> 02:42:38.477
this and ban that, restrict this and do
that. And then there's this like other

02:42:38.560 --> 02:42:41.517
like force out there that like doesn't
believe we have any power over them

02:42:41.600 --> 02:42:44.157
whatsoever. >> Uh, and they
have no desire to sign up

02:42:44.240 --> 02:42:47.117
for whatever rules we decide to put
in place. Um, and they're going to do

02:42:47.200 --> 02:42:50.957
whatever it is they're going to do
and we have no control over it at all. And

02:42:51.040 --> 02:42:52.797
it's China and specifically the Chinese

02:42:52.880 --> 02:42:55.517
Communist Party. Um, and they have a

02:42:55.600 --> 02:42:58.957
completely publicized open, you know,

02:42:59.040 --> 02:43:03.037
plan for what they're going to do with AI.
And it is not what we have in mind.

02:43:03.120 --> 02:43:06.397
Um, and not only do they have that as a
vision and a plan for their society, but

02:43:06.480 --> 02:43:08.717
they also have it as a vision
and plan for the rest of the world.

02:43:08.800 --> 02:43:10.557
>> So, their plan is what? Surveillance.

02:43:10.640 --> 02:43:14.797
>> Yeah, authoritarian control. So,
authoritarian population control. Um,

02:43:14.880 --> 02:43:18.557
you know, good good oldfashioned
communist authoritarian control. Um, and

02:43:18.640 --> 02:43:22.637
surveillance and enforcement. Um, and
social credit scores and all the rest of

02:43:22.720 --> 02:43:26.637
it. Um, and you are going to be
monitored and metered within an inch of

02:43:26.720 --> 02:43:30.077
everything all the time. Um, and it's,
you know, it's basically the end of

02:43:30.160 --> 02:43:31.677
human freedom. And that's their goal.

02:43:31.760 --> 02:43:34.637
and you know they justify it on the
basis of that's what leads to peace

02:43:34.720 --> 02:43:37.837
>> and you're worried that the uh

02:43:37.920 --> 02:43:42.797
regulating in the United States will
will halt progress enough to where uh

02:43:42.880 --> 02:43:44.957
the Chinese government
would win that race.

02:43:45.040 --> 02:43:47.997
>> So their plan Yeah. Yes. Yes.
And the reason for that is they and again

02:43:48.080 --> 02:43:50.557
they're very public on this. They they
have their plan is to proliferate their

02:43:50.640 --> 02:43:53.357
approach around the world. Um and they
have this program called the digital

02:43:53.440 --> 02:43:56.157
Silk Road, right, which is building
on their their Silk Road investment

02:43:56.240 --> 02:43:58.477
program. And they've got their they've
been laying they've been laying

02:43:58.560 --> 02:44:01.597
networking infrastructure all over
the world with their 5G, right, work with

02:44:01.680 --> 02:44:04.797
their company Huawei. And so they
they've been laying all this fabric,

02:44:04.880 --> 02:44:07.517
financial and technological fabric all
over the world. And their plan is to

02:44:07.600 --> 02:44:10.877
roll out their vision of AI on top
of that and to have every other country be

02:44:10.960 --> 02:44:15.197
running their version. And then if
you're a country prone to, you know,

02:44:15.280 --> 02:44:17.997
authoritarianism, you're going to find
this to be an incredible way to become

02:44:18.080 --> 02:44:21.117
more authoritarian. Uh [snorts] if
you're a country, by the way, not prone

02:44:21.200 --> 02:44:23.677
to authoritarianism, you're going to have
the Chinese Communist Party running

02:44:23.760 --> 02:44:27.357
your infrastructure and having back
doors into it, [laughter] right? Which

02:44:27.440 --> 02:44:30.717
is also not good. Um >> what's
your sense of where they stand in

02:44:30.800 --> 02:44:35.277
terms of the race towards uh super
intelligence as compared to the United

02:44:35.360 --> 02:44:37.117
States? >> Yeah.
So good news is they're behind,

02:44:37.200 --> 02:44:40.237
but bad news is they, you know,
they let's just say they get access to

02:44:40.320 --> 02:44:43.757
everything we do. Um, so they're
probably a year behind at each point in

02:44:43.840 --> 02:44:46.957
time, but they get, you know, downloads,
I think, of basically all of our work on

02:44:47.040 --> 02:44:51.357
a regular basis through a variety
of means. Um, and they are, you know, at

02:44:51.440 --> 02:44:53.677
least, we'll see they're at least
putting out reports of very, they just

02:44:53.760 --> 02:44:59.037
put out a report last week of a of a
GPT2 3.5 analog. um they put out this

02:44:59.120 --> 02:45:01.837
report. I forget what it's called,
but um they put out this report of this LM

02:45:01.920 --> 02:45:04.717
they did and they they you know the way
when OpenAI you know puts out they they

02:45:04.800 --> 02:45:08.877
they one of the ways they test you know
GPT um is they they they run it through

02:45:08.960 --> 02:45:12.477
standardized exams like the SAT right
how you can kind of gauge how smart it

02:45:12.560 --> 02:45:17.277
is. Uh, and so the Chinese report, they
ran their LLM through, uh, the Chinese

02:45:17.360 --> 02:45:21.837
equivalent of the SAT. Um,
and it includes a section on Marxism,

02:45:21.920 --> 02:45:25.117
>> um, and a section on I was say tongue
and thought. And it turns out their AI

02:45:25.200 --> 02:45:26.637
does very well on both of those topics.

02:45:26.720 --> 02:45:30.397
>> Oh, >> right. [laughter] So like,

02:45:30.480 --> 02:45:33.757
>> uh, this this alignment thing,
>> communist AI, right? Like literal

02:45:33.840 --> 02:45:38.397
communist AI, right? And so their vision
is like that's the you know so you know

02:45:38.480 --> 02:45:41.117
you can just imagine like you're
a school you know you're a kid 10 years

02:45:41.200 --> 02:45:44.477
from now in Argentina or in Germany or

02:45:44.560 --> 02:45:49.197
in who knows where uh Indonesia and you
ask the AI to explain to you like how

02:45:49.280 --> 02:45:52.317
the economy works and it gives you
the most cheery upbeat explanation of

02:45:52.400 --> 02:45:56.637
Chinese style communism you've ever
heard right so like the stakes here are

02:45:56.720 --> 02:45:59.837
like really big >> well
my as we've been talking about my

02:45:59.920 --> 02:46:03.517
hope is not just with the United States
but with just uh the kitten his basement

02:46:03.600 --> 02:46:06.077
with open source or SLM cuz I I don't

02:46:06.160 --> 02:46:09.517
know if I um trust large centralized

02:46:09.600 --> 02:46:12.557
institutions with super powerful AI no

02:46:12.640 --> 02:46:17.837
matter what their ideology
>> cuz uh power corrupts.

02:46:17.920 --> 02:46:19.517
You've been investing in tech companies

02:46:19.600 --> 02:46:22.637
for about let's say 20 years and uh

02:46:22.720 --> 02:46:25.357
about 15 of which was uh with Andre and

02:46:25.440 --> 02:46:30.477
Horowitz. Uh what interesting trends in
tech have you seen over that time? Let's

02:46:30.560 --> 02:46:33.677
just talk about companies and just
the evolution of the tech industry.

02:46:33.760 --> 02:46:37.597
>> I mean the big shift over 20 years
has been that tech used to be a tools

02:46:37.680 --> 02:46:42.797
industry. Uh for basically from like
1940 through to about 2010 almost all

02:46:42.880 --> 02:46:47.197
the big successful companies were pick
and shovels companies. So PC, database,

02:46:47.280 --> 02:46:50.557
smartphone, you know, some some some
tool that somebody else would pick up

02:46:50.640 --> 02:46:53.677
and use. Since 2010, most of the big

02:46:53.760 --> 02:46:56.477
wins have been in applications. Um so a

02:46:56.560 --> 02:47:00.797
company that starts a uh you know
starts in an existing industry and goes

02:47:00.880 --> 02:47:04.797
directly to the customer in that industry
and you know the early examples

02:47:04.880 --> 02:47:06.957
there were like Uber andyft and Airbnb.

02:47:07.040 --> 02:47:11.997
Um and then that model is kind
of elaborating out. Um uh the AI thing is

02:47:12.080 --> 02:47:14.557
actually a reversion on that for now
because like most of the AI business

02:47:14.640 --> 02:47:18.397
right now is actually in cloud provision
of of of AI APIs for other people to

02:47:18.480 --> 02:47:21.037
build on. But >> but the big
thing will probably be in app.

02:47:21.120 --> 02:47:24.877
>> Yeah. I think I think most of the money
I think probably will be in whatever.

02:47:24.960 --> 02:47:29.197
Yeah, you're AI financial adviser
or your AI doctor or your AI lawyer or you

02:47:29.280 --> 02:47:32.317
know take your pick of whatever
the domain is. Um and there and what's

02:47:32.400 --> 02:47:36.157
interesting is you know we the valley
kind of does everything. We we our

02:47:36.240 --> 02:47:39.277
entrepreneurs kind of elaborate every
possible idea. And so there will be a

02:47:39.360 --> 02:47:42.957
set of companies that like make AI um
something that can be purchased and used

02:47:43.040 --> 02:47:45.997
by large law firms. Um and then there
will be other companies that just go

02:47:46.080 --> 02:47:49.197
direct to market as a as an AI lawyer.

02:47:49.280 --> 02:47:52.317
What advice could you give for a startup

02:47:52.400 --> 02:47:56.797
founder? Just having seen so many
successful companies, so many companies

02:47:56.880 --> 02:48:00.637
that fail also. What advice could you
give to a startup founder, someone who

02:48:00.720 --> 02:48:03.357
wants to build the next super successful

02:48:03.440 --> 02:48:05.917
startup in the tech space, the Googles,

02:48:06.000 --> 02:48:10.397
the Apples, the Twitters. >> Yeah.
So, the great thing about the

02:48:10.480 --> 02:48:14.317
really great founders is they don't
take any advice. So, [laughter]

02:48:14.400 --> 02:48:18.797
so if you find yourself listening
to advice, maybe you shouldn't do it. Um

02:48:18.880 --> 02:48:23.357
>> but that's actually just to elaborate
on that if you could also speak to great

02:48:23.440 --> 02:48:27.677
founders too like what what makes
a great founder. So what makes a great

02:48:27.760 --> 02:48:30.237
founder is super smart um coupled with

02:48:30.320 --> 02:48:35.117
super energetic coupled with super
courageous. I think it's some of it's

02:48:35.200 --> 02:48:37.837
those three and >> intelligence,
passion and courage.

02:48:37.920 --> 02:48:41.357
>> The first two are traits and the third
one is a choice. I think courage is a

02:48:41.440 --> 02:48:43.037
choice. Well, because courage is a

02:48:43.120 --> 02:48:46.957
question of pain, tolerance, right? Um

02:48:47.040 --> 02:48:51.597
so um how how many times you're willing to
get punched in the face before you quit?

02:48:51.680 --> 02:48:54.877
>> Yeah. Um, and here's
maybe the biggest thing people

02:48:54.960 --> 02:48:57.917
don't understand about what it's like
to be a startup founder is it gets it gets

02:48:58.000 --> 02:49:00.797
very romanticized, right? Um, and even
when it even when they fail, it still

02:49:00.880 --> 02:49:04.477
gets romanticized about like what
a great adventure it was. But like the

02:49:04.560 --> 02:49:07.837
reality of it is most of what happens
is people telling you no. And then they

02:49:07.920 --> 02:49:09.517
usually follow that with you're stupid.

02:49:09.600 --> 02:49:11.437
>> Yeah. >> Right. No,
I will not come to work for

02:49:11.520 --> 02:49:14.397
you. Um, I will not leave my cushy job
at Google to come work for you. No, I'm

02:49:14.480 --> 02:49:16.877
not going to buy your product. You know,
no, I'm not going to run a story about

02:49:16.960 --> 02:49:20.877
your company. No, I'm not this, that,
the other thing. Um, and so a huge

02:49:20.960 --> 02:49:23.677
amount of what people have to do is just
get used to just getting punched. And

02:49:23.760 --> 02:49:26.157
and and and the reason people don't
understand this is because when you're a

02:49:26.240 --> 02:49:28.797
founder, you cannot let on that this is
happening because it will cause people

02:49:28.880 --> 02:49:30.957
to think that you're weak
and they'll lose faith in you.

02:49:31.040 --> 02:49:32.477
>> Yeah. >> So you have
to pretend that you're

02:49:32.560 --> 02:49:37.197
having a great time when you're
dying inside, right? [laughter]

02:49:37.280 --> 02:49:39.997
>> Just in misery.
>> But why why did they do it?

02:49:40.080 --> 02:49:41.517
>> Why did they do Yeah, that's the thing.

02:49:41.600 --> 02:49:44.317
It's it's like it is a level. This is
actually one of the conclusions I think

02:49:44.400 --> 02:49:47.517
is that I think it's actually for most
of these people on a riskadjusted basis

02:49:47.600 --> 02:49:49.277
that's probably an irrational act.

02:49:49.360 --> 02:49:51.997
>> They could probably be more financially
successful on average if they just got

02:49:52.080 --> 02:49:55.757
like a real job in at a big company.
Um but there's [snorts] you know some

02:49:55.840 --> 02:49:59.117
people just have an irrational need to do
something new and build something for

02:49:59.200 --> 02:50:02.237
themselves. And you know some people
just can't tolerate having bosses. Oh

02:50:02.320 --> 02:50:04.797
here's a fun thing is how do
you reference check founders?

02:50:04.880 --> 02:50:06.157
>> Right? So you call,
you know, normal way

02:50:06.240 --> 02:50:09.837
you reference Jackie hiring somebody is
you call the bosses and you know and you

02:50:09.920 --> 02:50:12.397
find out if they were good employees
and now you're trying to reference check

02:50:12.480 --> 02:50:15.517
Steve Jobs, right? And it's like, oh god,
he was terrible, you know, he was a

02:50:15.600 --> 02:50:17.437
terrible employee. He never
did what we told him to do.

02:50:17.520 --> 02:50:21.837
>> Yeah. [laughter] >> So
what's a good reference?

02:50:21.920 --> 02:50:25.517
If you want the previous boss to actually
say they're they never did what

02:50:25.600 --> 02:50:27.437
you told them to do,
that might be a good thing.

02:50:27.520 --> 02:50:30.877
>> Well, ideally ideally what you want is
I will go I I would like to go to work for

02:50:30.960 --> 02:50:34.717
that person. um he worked for me here
and now I'd like to work for him. Now

02:50:34.800 --> 02:50:38.077
unfortunately most people can't
their egos can't can't handle that so they

02:50:38.160 --> 02:50:39.917
won't say that but that
that that's the ideal.

02:50:40.000 --> 02:50:43.437
>> What advice would you give to those
folks in the space of intelligence,

02:50:43.520 --> 02:50:48.317
passion and courage? So I think
the other big thing is you see people

02:50:48.400 --> 02:50:50.717
sometimes who say I want to start
a company and then they kind of work

02:50:50.800 --> 02:50:54.397
through the process of coming up with
an idea and generally those don't work as

02:50:54.480 --> 02:50:58.397
well as the case where somebody has the
idea first and then they kind of realize

02:50:58.480 --> 02:51:01.037
that there's an opportunity to build
a company and then they just turn out to

02:51:01.120 --> 02:51:02.637
be the right kind of person to do that.

02:51:02.720 --> 02:51:07.197
When you say idea, do you mean >>
long-term big vision or do you mean

02:51:07.280 --> 02:51:11.757
specifics of like product?
>> Spec I say specific like specifically

02:51:11.840 --> 02:51:14.637
what yes specifics like what is the
because for the first five years you

02:51:14.720 --> 02:51:16.957
don't get to have vision. You just got
to build something people want and you

02:51:17.040 --> 02:51:19.997
got to figure out a way to sell it
to them, right? It's very practical or you

02:51:20.080 --> 02:51:22.877
never get to big vision. So >> So
the first the first pro you you have

02:51:22.960 --> 02:51:26.317
an idea of set of products or the first
product that can actually make some money.

02:51:26.400 --> 02:51:28.877
>> Yeah. Like it's got to the first
product's got to work. By which I mean

02:51:28.960 --> 02:51:32.157
like it has to technically work but then
it has to actually fit into the category

02:51:32.240 --> 02:51:34.637
in the customer's mind of something that
they want and then and then by the way

02:51:34.720 --> 02:51:36.877
the other part is they have to willing
to pay for it. Like somebody's got to

02:51:36.960 --> 02:51:39.837
pay the bills and so you got to figure
out how to price it and whether you can

02:51:39.920 --> 02:51:41.437
actually extract the money. >> Yeah.

02:51:41.520 --> 02:51:46.317
>> So usually it is much more predictable.

02:51:46.400 --> 02:51:49.117
Success is never predictable but it's
more predictable if you start with a

02:51:49.200 --> 02:51:52.957
great idea and then back into starting
the company. Um so this is what we did.

02:51:53.040 --> 02:51:55.597
You know we had Mosaic before we had
Netscape. the Google guys had the Google

02:51:55.680 --> 02:51:59.757
search engine working at Stanford. Um,
right. Um, the um, uh, you know, yeah,

02:51:59.840 --> 02:52:03.037
actually there's tons of examples where
they, you know, Pierre Omdar had eBay

02:52:03.120 --> 02:52:05.037
working before he left his previous job.

02:52:05.120 --> 02:52:08.557
>> So, I really love that idea of just
having a thing, a prototype that

02:52:08.640 --> 02:52:11.757
actually works before you
even begin to remotely scale.

02:52:11.840 --> 02:52:14.717
>> Yeah. By the way, it's also far
easier to raise money, right? Like the the

02:52:14.800 --> 02:52:17.517
ideal pitch that we receive is here's
a thing that works. Would you like to

02:52:17.600 --> 02:52:21.277
invest in our company or not? Like that's
so much easier than here's 30 slides

02:52:21.360 --> 02:52:25.117
>> with a dream, right? [clears throat] Um
and then we have this concept called the

02:52:25.200 --> 02:52:28.717
idea maze which our biology friend
of Boston came up with um when he was with

02:52:28.800 --> 02:52:33.517
us. Um so so so then there's this thing
this goes to mythology which is um you

02:52:33.600 --> 02:52:36.717
know there's a mythology that kind
of you know these these ideas um you know

02:52:36.800 --> 02:52:39.437
kind of arrive like magic or people
kind of stumble into them. It's like eBay

02:52:39.520 --> 02:52:41.357
with the pest dispensers or something.

02:52:41.440 --> 02:52:44.077
Um the reality usually with the big

02:52:44.160 --> 02:52:48.397
successes is that the founder has been
chewing on the problem for five or 10

02:52:48.480 --> 02:52:53.117
years before they start the company
and they often worked on it in school.

02:52:53.200 --> 02:52:57.037
>> Um or they even experimented on it when
they were a kid um and they've been kind

02:52:57.120 --> 02:53:00.157
of training up over that period of time
to be able to do the thing. So they're

02:53:00.240 --> 02:53:04.477
like a true domain expert and and and it
sort of sounds like mom and apple pie

02:53:04.560 --> 02:53:06.477
which is yeah you want to be a domain
expert in what you're doing but you

02:53:06.560 --> 02:53:09.837
would you know the mythology is so
strong of like oh I just like had this

02:53:09.920 --> 02:53:11.597
idea in the shower
and now I'm doing it like

02:53:11.680 --> 02:53:16.237
>> it's generally not that. No, because
well maybe in the shower we had the

02:53:16.320 --> 02:53:21.197
exact product implementation details, but

02:53:21.280 --> 02:53:22.637
yeah, usually you're going to be for

02:53:22.720 --> 02:53:26.877
like years if not decades thinking about

02:53:26.960 --> 02:53:31.597
like everything around that.
>> Well, we call it the idea maze because

02:53:31.680 --> 02:53:35.117
the the idea maze basically is like
there's all these permutations like for

02:53:35.200 --> 02:53:37.917
any ide for any idea there's like all
these different permutations. Who should

02:53:38.000 --> 02:53:40.557
the customer be? What shape form should
the product have and how should we take

02:53:40.640 --> 02:53:45.597
it to market and all these things. Um
and so um the really smart founders have

02:53:45.680 --> 02:53:48.877
thought through all these scenarios
by the time they go out to raise money. Um

02:53:48.960 --> 02:53:52.717
and they have like detailed answers
um on every one of those fronts because

02:53:52.800 --> 02:53:54.397
they put so much thought into it.

02:53:54.480 --> 02:53:58.637
>> Um the sort of the the the sort of more
haphazard founders haven't thought about

02:53:58.720 --> 02:54:01.757
any any of that and it's the detailed
ones who tend to do much better. So

02:54:01.840 --> 02:54:06.557
>> how do you know when to take a leap
if you have a cushy job or happy life?

02:54:06.640 --> 02:54:09.837
>> I mean the best reason is just cuz you
can't tolerate not doing it, right? Like

02:54:09.920 --> 02:54:12.077
this is the kind of thing where if you
have to be advised into doing it, you

02:54:12.160 --> 02:54:15.597
probably shouldn't do it. Um, and so
it's probably the opposite, which is you

02:54:15.680 --> 02:54:18.797
just have such a burning sense of this
has to be done. I have to do this. I

02:54:18.880 --> 02:54:21.677
have no choice. >> What if it's
going to lead to a lot of pain?

02:54:21.760 --> 02:54:23.007
>> It's going to lead to a lot of pain.

02:54:23.090 --> 02:54:24.637
[laughter] I think that's

02:54:24.720 --> 02:54:26.797
>> what if it means uh losing sort of

02:54:26.880 --> 02:54:32.077
social relationships and damaging your um
relationship with loved ones and all

02:54:32.160 --> 02:54:34.637
that kind of stuff. >> Yeah.
Look, so like it's going to put

02:54:34.720 --> 02:54:36.237
you in a social tunnel for sure, right?

02:54:36.320 --> 02:54:39.757
So you're going to like I you know
there's this game you can play on

02:54:39.840 --> 02:54:43.357
Twitter which is you can do any whiff
of the idea that there's uh basically any

02:54:43.440 --> 02:54:46.477
such thing as work life balance and
that people should actually work hard and

02:54:46.560 --> 02:54:49.197
everybody gets mad. But like the truth
is like all the successful founders are

02:54:49.280 --> 02:54:53.037
working 80our weeks and they're working
you know they form very very strong

02:54:53.120 --> 02:54:55.517
social bonds with the people they work
with. They tend to lose a lot of friends

02:54:55.600 --> 02:54:58.637
on the outside or put those friendships
on ice. Like that's just the nature of

02:54:58.720 --> 02:55:02.797
the of the thing. Um you know for most
people that's worth the trade-off. you

02:55:02.880 --> 02:55:05.197
know, the advantage, you know, maybe
younger founders have is maybe they have

02:55:05.280 --> 02:55:07.277
less, you know, maybe they're not,
you know, for example, if they're not

02:55:07.360 --> 02:55:10.157
married yet or don't have kids yet,
that's an easier thing to bite off.

02:55:10.240 --> 02:55:11.437
>> Can you be an older founder?

02:55:11.520 --> 02:55:14.477
>> Yeah, you definitely can. Yeah.
Um, yeah, many of the most successful

02:55:14.560 --> 02:55:17.197
founders are second, third, fourth time
founders. They're in their 30s, 40s,

02:55:17.280 --> 02:55:21.117
50s. Um, the good news with being
an older founder is you know more and you

02:55:21.200 --> 02:55:23.837
you know a lot more about what to do,
which is very helpful. The problem is,

02:55:23.920 --> 02:55:27.437
okay, now you've got like a spouse and
a family and kids and like you've got to

02:55:27.520 --> 02:55:31.677
go to the baseball game and like you can't
go to the baseball, you know, and so it's

02:55:31.760 --> 02:55:33.917
>> life is full of difficult choices. Yes.

02:55:34.000 --> 02:55:37.437
And uh >> you've written
a blog post on what

02:55:37.520 --> 02:55:39.357
you've been up to. Uh you wrote this in

02:55:39.440 --> 02:55:43.037
October 2022. Uh quote, mostly I try to

02:55:43.120 --> 02:55:48.157
learn a lot. For example, the political
events of 2014 to 2016 made clear to me

02:55:48.240 --> 02:55:49.837
that I didn't understand politics at

02:55:49.920 --> 02:55:52.637
all. referencing maybe some of this this

02:55:52.720 --> 02:55:55.277
book here. Um, so I deliberately

02:55:55.360 --> 02:55:56.957
withdrew from political engagement and

02:55:57.040 --> 02:55:59.677
fundraising and instead read my way back

02:55:59.760 --> 02:56:02.717
into history and as far to the political

02:56:02.800 --> 02:56:07.117
left and political right as I could.
So just high level question, what's your

02:56:07.200 --> 02:56:09.037
approach to learning?

02:56:09.120 --> 02:56:10.957
>> Yeah.
So it's basically I would say it's

02:56:11.040 --> 02:56:14.157
it's autodidact. Um, uh, so it's sort of

02:56:14.240 --> 02:56:16.157
goes it's going down the rabbit holes.

02:56:16.240 --> 02:56:19.597
Um, so it's a combination. So I kind of
allude to it in that in that quote. It's

02:56:19.680 --> 02:56:21.517
a combination of breadth and depth.

02:56:21.600 --> 02:56:25.357
>> Um, and so I tend to Yeah, I tend to I
I go broad by the nature of what I do. I

02:56:25.440 --> 02:56:27.917
go broad, but then I tend to go deep
in a rabbit hole for a while, read

02:56:28.000 --> 02:56:30.797
everything I can and then come out of it
and I might I might not revisit that

02:56:30.880 --> 02:56:32.717
rabbit hole for, you know, another decade.

02:56:32.800 --> 02:56:37.277
>> And in that blog post that I recommend
people go check out, you actually list a

02:56:37.360 --> 02:56:40.717
bunch of different books that you
recommend on different topics on the

02:56:40.800 --> 02:56:43.037
American left and the American right.

02:56:43.120 --> 02:56:46.797
Uh, it's just a lot of really good stuff.
The best explanation for the

02:56:46.880 --> 02:56:49.917
current structure of our society and
politics. You give two recommendations.

02:56:50.000 --> 02:56:53.517
Four books on the Spanish Civil War.
Six books on deep history of the American

02:56:53.600 --> 02:56:56.157
right. Comprehensive biocracy of Adolf

02:56:56.240 --> 02:57:01.197
Hitler. U one of which I read can
recommend. Uh six books on the deep

02:57:01.280 --> 02:57:04.397
history of the American left.
So American right, American left, looking

02:57:04.480 --> 02:57:06.637
at the history to give you the context.

02:57:06.720 --> 02:57:10.317
Um biography of uh Vlad Lennon, two of

02:57:10.400 --> 02:57:13.837
them uh on the French Revolution.
I actually have never read a biography on

02:57:13.920 --> 02:57:16.317
Lenin. Maybe that that would be useful.

02:57:16.400 --> 02:57:18.477
Everything's been so Marx focused.

02:57:18.560 --> 02:57:21.197
>> The Sebastian biography
of Lenin is extraordinary.

02:57:21.280 --> 02:57:23.677
>> Uh Victor Sebastian
would blow your mind. Yeah.

02:57:23.760 --> 02:57:24.877
>> So it's still useful to read.

02:57:24.960 --> 02:57:25.917
>> It's incredible. Yeah. It's incredible.

02:57:26.000 --> 02:57:28.557
I actually think it's the single
best book on the Soviet Union.

02:57:28.640 --> 02:57:32.077
>> So that the perspective of Lenin
is might be the best way to look at the

02:57:32.160 --> 02:57:34.157
Soviet Union versus Stalin versus Marx

02:57:34.240 --> 02:57:36.797
versus Very interesting. So two books on

02:57:36.880 --> 02:57:39.597
fascism and anti-fascism

02:57:39.680 --> 02:57:42.797
uh by the same uh author Paul Gotthrey.

02:57:42.880 --> 02:57:46.477
Uh brilliant book on the nature of mass
movements and collective psychology. The

02:57:46.560 --> 02:57:50.797
definitive work on intellectual life
under totalitarianism, the captive mind.

02:57:50.880 --> 02:57:54.957
Uh the definitive work on the practical
life under totalitarianism. Uh there's a

02:57:55.040 --> 02:57:59.197
bunch there's a bunch. And the single
best book first of all the list here is

02:57:59.280 --> 02:58:01.677
just incredible. But you say the single

02:58:01.760 --> 02:58:04.237
best book I have found on who we are and

02:58:04.320 --> 02:58:06.557
how we got here is the ancient city uh

02:58:06.640 --> 02:58:10.957
by Numad Dennis Fell Kulangis.

02:58:11.040 --> 02:58:15.277
I like it. uh what's uh what did
you learn about who we are as a human

02:58:15.360 --> 02:58:18.077
civilization from that book?
>> Yeah, so this is a fascinating book.

02:58:18.160 --> 02:58:19.437
This one's free. It's a free by the way.

02:58:19.520 --> 02:58:22.797
It's it's a book from the 1860s. You can
download it or you can buy print outs of

02:58:22.880 --> 02:58:26.797
prints of it. But um it's uh it was this
guy who was a professor at the Sarbon in

02:58:26.880 --> 02:58:29.037
the 1860s. And he was apparently a

02:58:29.120 --> 02:58:31.917
savant on uh antiquity on on Greek and

02:58:32.000 --> 02:58:36.397
Roman antiquity. Um and and the reason I
say that is because his sources are 100%

02:58:36.480 --> 02:58:40.717
original Greek and Roman sources. So he
wrote a basically a history of western

02:58:40.800 --> 02:58:44.477
civilization from on the order of 4,000
years ago to basically the present times

02:58:44.560 --> 02:58:47.037
entirely working on original Greek and

02:58:47.120 --> 02:58:51.917
and Roman Roman sources. Um and what he
was specifically trying to do was he was

02:58:52.000 --> 02:58:54.557
trying to reconstruct from the stories
of the Greeks and the Romans. He was

02:58:54.640 --> 02:58:57.597
trying to reconstruct what life in the
west was like before the Greeks and the

02:58:57.680 --> 02:59:00.797
Romans which was in this in this
in the civilization known as the the

02:59:00.880 --> 02:59:04.397
Indo-Europeans. Um, and the short answer

02:59:04.480 --> 02:59:09.437
is, and this is sort of circa 4,000, you
know, 2,000 BC to, you know, sort of 500

02:59:09.520 --> 02:59:13.037
BC, kind of that 1500 year stretch
where civilization developed. Uh, and his

02:59:13.120 --> 02:59:15.997
conclusion was basically cults. Um, they

02:59:16.080 --> 02:59:20.397
were basically cults. And civilization
was or organized into cults and the the

02:59:20.480 --> 02:59:24.477
intensity of the cults was like
a millionfold beyond anything that we

02:59:24.560 --> 02:59:26.157
would recognize today. like it was a

02:59:26.240 --> 02:59:29.277
level of um all-encompassing

02:59:29.360 --> 02:59:33.197
belief and uh an action around religion

02:59:33.280 --> 02:59:36.717
um that was at a level of extremistness
that we we wouldn't even recognize it.

02:59:36.800 --> 02:59:41.517
Um uh and and so specifically he tells
the story of basically there were three

02:59:41.600 --> 02:59:44.957
levels of cults. There was the family
cult, the tribal cult, and then the city

02:59:45.040 --> 02:59:48.397
cult as as society scaled up. And then

02:59:48.480 --> 02:59:51.677
each cult was a joint cult of uh family

02:59:51.760 --> 02:59:53.917
gods which were ancestor gods and then

02:59:54.000 --> 02:59:59.277
nature gods. Um and then your bonding
into a family, a tribe or a city was

02:59:59.360 --> 03:00:01.517
based on your adherence to that religion.

03:00:01.600 --> 03:00:06.077
>> Um people uh who were not of your
family, tribe, city worship different

03:00:06.160 --> 03:00:11.477
gods which gave you not just the right but
the responsibility to kill them on site.

03:00:11.560 --> 03:00:14.557
[laughter] >> Right. >> So they
were serious about their cults

03:00:14.640 --> 03:00:18.157
>> hardcore. By the way, shocking
development. I did not realize there's

03:00:18.240 --> 03:00:21.757
zero concept of individual rights. Like
even even up through the Greeks and even

03:00:21.840 --> 03:00:24.477
in the Romans, they didn't have have
the concept of individual rights. Like the

03:00:24.560 --> 03:00:27.997
idea that as an individual you have like
some right it's just like nope, right?

03:00:28.080 --> 03:00:30.637
And you look back and you're just
like wow that's just like crazily like

03:00:30.720 --> 03:00:33.677
fascist in a degree that we wouldn't
recognize today. But it's like well they

03:00:33.760 --> 03:00:37.837
were living under extreme pressure for
survival and you and you know the theory

03:00:37.920 --> 03:00:40.157
goes you could not have people running
around making claims to individual

03:00:40.240 --> 03:00:42.237
rights when you're just trying to get
like your tribe through the winter,

03:00:42.320 --> 03:00:46.077
right? Like you need like hardcore
command and control. And so and and and

03:00:46.160 --> 03:00:48.557
actually what what if through a modern
political lens those cults were

03:00:48.640 --> 03:00:52.157
basically both fascist and communist
um they were fascist in terms of social

03:00:52.240 --> 03:00:55.357
control and then they were
communist in terms of economics.

03:00:55.440 --> 03:00:57.437
[clears throat] >> But you
think that's fundamentally that

03:00:57.520 --> 03:01:01.277
like pull towards uh cults is within us.

03:01:01.360 --> 03:01:04.237
>> Well so so my conclusion from this book

03:01:04.320 --> 03:01:08.797
so so so the way we naturally think
about the world we live in today is like

03:01:08.880 --> 03:01:11.997
we basically have such an improved
version of everything that came before

03:01:12.080 --> 03:01:14.797
us, right? like we we we have basically
we've figured out all these things

03:01:14.880 --> 03:01:17.517
around morality and ethics and democracy
and all these things and like they were

03:01:17.600 --> 03:01:20.717
basically stupid and retrograde
and we're like smart and sophisticated and

03:01:20.800 --> 03:01:24.717
we've improved all this. Um I I after
reading that book uh I I now believe in

03:01:24.800 --> 03:01:27.837
many ways the opposite which is no
actually we are still running in that

03:01:27.920 --> 03:01:32.237
original model. We're just running
in an incredibly diluted version of it. So

03:01:32.320 --> 03:01:34.237
we're still running basically in cults.

03:01:34.320 --> 03:01:37.037
It's just our cults are at like
a thousandth or a millionth the level of

03:01:37.120 --> 03:01:41.437
intensity. Right? And so our so just
to take religions you know the modern

03:01:41.520 --> 03:01:45.357
experience of a Christian in our time
even somebody who considers him a devout

03:01:45.440 --> 03:01:48.717
Christian is just a shadow of the level
of intensity of somebody who belonged to

03:01:48.800 --> 03:01:52.717
a religion back in that period >>
and then by the way we have con it goes

03:01:52.800 --> 03:01:57.037
back to our AI discussion we we we we
then sort of endlessly create new cults

03:01:57.120 --> 03:01:59.277
like we're trying to fill the void right

03:01:59.360 --> 03:02:04.237
and the void is a void of of bonding
okay living in their era like everybody

03:02:04.320 --> 03:02:06.797
living today transported in that era
would view it as just completely

03:02:06.880 --> 03:02:09.757
intolerable in terms like the the loss
of freedom and the level of basically

03:02:09.840 --> 03:02:13.437
fascist control. However, every single
person in that era, and he really

03:02:13.520 --> 03:02:15.757
stresses this, they knew
exactly where they stood.

03:02:15.840 --> 03:02:17.277
>> They knew exactly where they belonged.

03:02:17.360 --> 03:02:19.757
They knew exactly what their purpose was.
They knew exactly what they needed

03:02:19.840 --> 03:02:22.237
to do every day. They knew exactly
why they were doing it. They had total

03:02:22.320 --> 03:02:24.077
certainty about their
place in the universe.

03:02:24.160 --> 03:02:27.357
>> So, the question of meaning and the
question of purpose was very distinctly

03:02:27.440 --> 03:02:30.877
clearly defined for them.
>> Absolutely. Overwhelmingly,

03:02:30.960 --> 03:02:35.517
undisputably, undeniably.
As we turn the volume down on the cultism,

03:02:35.600 --> 03:02:39.437
>> yes, >> we start to uh the search for
meaning starts getting harder and harder.

03:02:39.520 --> 03:02:42.477
>> Yes. Because we we don't have that.
We are we are ungrounded. We are we we are

03:02:42.560 --> 03:02:44.237
unentered and and we all feel it, right?

03:02:44.320 --> 03:02:47.677
And that's why we reach for, you know,
it's why we still reach for religion.

03:02:47.760 --> 03:02:50.797
It's why we reach for, you know, we
people start to take on, you know, let's

03:02:50.880 --> 03:02:53.997
say, you know, a faith in science maybe
beyond where they should put it. Uh, you

03:02:54.080 --> 03:02:56.557
know, and by the way, like sports teams
are like a, you know, they're like a

03:02:56.640 --> 03:02:59.837
tiny little version of a cult. And,
you know, you know, Apple keynotes are a

03:02:59.920 --> 03:03:02.477
tiny little version of a cult, right?

03:03:02.560 --> 03:03:04.477
and you know political
you know >> yeah [snorts]

03:03:04.560 --> 03:03:06.557
>> and there's cult you know there's
full-blown cults on both sides of the

03:03:06.640 --> 03:03:09.357
political spectrum right now right
um you know operating in plain sight

03:03:09.440 --> 03:03:11.757
>> but still not full-blown
compared as to what it was

03:03:11.840 --> 03:03:14.717
>> compared to what it used to I mean we
would today consider full-blown but like

03:03:14.800 --> 03:03:17.437
yes they're they're at like I don't
know a 100 thousandth or something of the

03:03:17.520 --> 03:03:21.197
intensity of of what people had back
then so so we live in a world today that

03:03:21.280 --> 03:03:23.917
in many ways is more advanced and moral
and so forth and it's certainly a lot

03:03:24.000 --> 03:03:27.277
nicer much nicer world to live in but we
live in a world that's like very washed

03:03:27.360 --> 03:03:31.597
out it's like everything has become very
colorless and gray as compared to how

03:03:31.680 --> 03:03:34.637
people used to experience things, which
is I think why we're so prone to reach

03:03:34.720 --> 03:03:38.237
for drama. We we there's something in us

03:03:38.320 --> 03:03:43.037
deeply evolved where we want that back
>> and I wonder where it's all headed

03:03:43.120 --> 03:03:46.557
>> as we turn the volume down more
and more. Uh what advice would you give to

03:03:46.640 --> 03:03:50.717
young folks today [snorts] uh in high
school and college, how to be successful

03:03:50.800 --> 03:03:53.037
in their career, how to be
successful in their life?

03:03:53.120 --> 03:03:58.077
>> Yes. So the tools that are
available today, I mean, are just like I

03:03:58.160 --> 03:04:01.037
sometimes, you know, bore I sometimes
bore uh, you know, kids by describing

03:04:01.120 --> 03:04:03.677
like what it was like to go look up
a book, you know, to try to like discover

03:04:03.760 --> 03:04:07.117
a fact in, you know, in in the old days,
the 1970s, 1980s and go to the library

03:04:07.200 --> 03:04:09.597
and the card catalog and the whole thing.
you go through all that work and

03:04:09.680 --> 03:04:13.197
then the book is checked out and you
have to wait two weeks and like like to

03:04:13.280 --> 03:04:16.637
be in a world not only where you can get
the answer to any question but also the

03:04:16.720 --> 03:04:19.437
world now you know the AI world where
you've got like the assistant that will

03:04:19.520 --> 03:04:22.477
help you do anything help you teach
learn anything like your ability both to

03:04:22.560 --> 03:04:26.397
learn and also to produce is just like
I don't know a millionfold beyond what it

03:04:26.480 --> 03:04:30.237
used to be I have a I have a blog post
I've been wanting to write um which I

03:04:30.320 --> 03:04:32.477
call where where are the hyperproductive

03:04:32.560 --> 03:04:37.997
people um like good question >>
right like with these tools like there

03:04:38.080 --> 03:04:40.557
should be authors that are writing
like hundreds or thousands of like

03:04:40.640 --> 03:04:43.437
outstanding books. >> Well,
with the authors, there's a

03:04:43.520 --> 03:04:47.837
consumption question, too. But yeah,
well, maybe not. Maybe not. You're

03:04:47.920 --> 03:04:51.677
right. But so the tools are much
more powerful, getting much more

03:04:51.760 --> 03:04:55.517
>> artist musicians, right? Why aren't
musicians producing a thousand times the

03:04:55.600 --> 03:04:58.157
number of songs, right? Um like the

03:04:58.240 --> 03:04:59.997
tools are spectacular.

03:05:00.080 --> 03:05:02.637
>> So what uh what's the explanation? And

03:05:02.720 --> 03:05:08.077
by way of advice, like mo is motivation
starting to be turned down a little bit

03:05:08.160 --> 03:05:10.397
or what? >> I think it
might be distraction.

03:05:10.480 --> 03:05:12.637
>> Distraction.
>> It's it's so easy to just sit and

03:05:12.720 --> 03:05:17.117
consume um that I think people get
distracted from production. But if you

03:05:17.200 --> 03:05:19.997
wanted to um you know, as a young
person, if you wanted to really stand

03:05:20.080 --> 03:05:24.957
out, you could get on a like a a hyper
productivity curve very early on.

03:05:25.040 --> 03:05:27.437
There's a great uh you know the story
there's a great story in Roman history

03:05:27.520 --> 03:05:31.677
of Plenty the Elder who was this
legendary statesman um died in the

03:05:31.760 --> 03:05:35.117
Vuvius eruption trying to rescue his
friends but um he was famous both for

03:05:35.200 --> 03:05:38.557
being a a savant basically being
a polymath but also being an author and he

03:05:38.640 --> 03:05:41.117
wrote apparently like hundreds of books
most of which have been lost but he like

03:05:41.200 --> 03:05:44.477
wrote all these encyclopedias
and [snorts] he literally like would be

03:05:44.560 --> 03:05:47.597
reading and writing all day long no
matter what else was going on and he so

03:05:47.680 --> 03:05:50.237
he would like travel with like four
slaves and two of them were responsible

03:05:50.320 --> 03:05:54.157
for reading to him and two of them were
responsible for taking dictation and So

03:05:54.240 --> 03:05:57.037
like he'd be going cross country and like
literally he would be writing books

03:05:57.120 --> 03:06:00.157
like all the time and apparently they
were spectacular. There's only a few

03:06:00.240 --> 03:06:01.837
that have survived
but apparently they were amazing.

03:06:01.920 --> 03:06:05.277
>> So there's a lot of value to being
somebody who finds focus in this life.

03:06:05.360 --> 03:06:08.157
>> Yeah. Like when and there are examples
like there are uh you know there's this

03:06:08.240 --> 03:06:12.157
guy uh judge what's his name? Posner
Pner um who wrote like 40 books and was

03:06:12.240 --> 03:06:16.157
also a great federal judge. Um you know
there's our friend Bali I think is like

03:06:16.240 --> 03:06:19.597
this. He's one of these you know where
he's his output is just prodigious. Um

03:06:19.680 --> 03:06:23.357
and so it's like yeah I mean with these
tools why not? And I kind of think we're

03:06:23.440 --> 03:06:26.157
we're at this interesting kind of freeze
frame moment where like this these tools

03:06:26.240 --> 03:06:28.157
are now in everybody's hands and
everybody's just kind of staring at them

03:06:28.240 --> 03:06:30.237
trying to figure out what to do.
The new tools.

03:06:30.320 --> 03:06:32.157
>> We have discovered fire. >> Yeah.

03:06:32.240 --> 03:06:34.317
>> And trying to figure
out how to use it to cook.

03:06:34.400 --> 03:06:38.157
>> Yeah. Right. >> Uh you told
Tim Ferrris that the perfect

03:06:38.240 --> 03:06:43.277
day is caffeine for 10 hours and alcohol
for 4 hours. You didn't think I'd be

03:06:43.360 --> 03:06:47.357
mentioning this, did you?
>> Uh it balances everything out perfectly

03:06:47.440 --> 03:06:50.237
as you said. >> So perfect.
Uh what's the So let me ask

03:06:50.320 --> 03:06:54.317
what's what's the secret to balance
and maybe to happiness in life.

03:06:54.400 --> 03:06:57.917
>> Um I I don't believe in balance.
So I I'm the wrong person to ask.

03:06:58.000 --> 03:07:00.237
>> Can you elaborate why
you don't believe in balance?

03:07:00.320 --> 03:07:04.157
>> I mean I I maybe it's just and I I look
I think people I think people are wired

03:07:04.240 --> 03:07:06.957
differently. So I I think it's hard
to generalize uh this kind of thing. But

03:07:07.040 --> 03:07:10.077
I'm I am much happier and more satisfied
when I'm fully committed to something.

03:07:10.160 --> 03:07:14.157
So I'm [snorts] very much
in favor of of imbalance. Yeah.

03:07:14.240 --> 03:07:19.037
>> Imbalance. And that applies to work,
to life, to everything. Yeah. No, no, I

03:07:19.120 --> 03:07:21.597
happen to have whatever twist
of personality traits lead that in

03:07:21.680 --> 03:07:24.957
non-destructive dimensions, including
the fact that I've actually I now no

03:07:25.040 --> 03:07:28.397
longer do the 104 plan. I I stopped
drinking. I do the caffeine, but not the

03:07:28.480 --> 03:07:31.917
alcohol. So, there's something
in my personality where I I I whatever

03:07:32.000 --> 03:07:34.797
maladaption I have is inclining
me towards productive things, not

03:07:34.880 --> 03:07:37.757
unproductive things. >> So,
you're one of the wealthiest people

03:07:37.840 --> 03:07:42.317
in the world. What's the relationship
between wealth and happiness?

03:07:42.400 --> 03:07:45.677
>> Oh, uh >> money
and happiness. So I think

03:07:45.760 --> 03:07:50.237
happiness I don't think
happiness is the thing

03:07:50.320 --> 03:07:53.597
>> to strive for.
>> I think satisfaction is the thing.

03:07:53.680 --> 03:07:56.797
>> That's that just sounds like
happiness but turned down a bit.

03:07:56.880 --> 03:08:01.517
>> No deeper. So happiness is you know
a walk in the woods at sunset. An ice

03:08:01.600 --> 03:08:02.957
cream cone. >> Mhm.

03:08:03.040 --> 03:08:05.917
>> A kiss. Um the first ice cream cone is

03:08:06.000 --> 03:08:10.317
great. The thousandth ice cream cone not
so much. At some point the walks in the

03:08:10.400 --> 03:08:14.397
woods get boring. What's the distinction
between happiness and uh satisfaction?

03:08:14.480 --> 03:08:16.877
>> I think satisfaction is a deeper thing

03:08:16.960 --> 03:08:18.717
which is like having found a purpose and

03:08:18.800 --> 03:08:22.317
fulfilling it being useful. So just uh

03:08:22.400 --> 03:08:24.477
uh something that permeates all your

03:08:24.560 --> 03:08:28.157
days just this general contentment of of

03:08:28.240 --> 03:08:31.757
being useful >> that I'm
fully satisfying my faculties

03:08:31.840 --> 03:08:35.437
that I'm fully delivering right uh
on the gifts that I've been given that I'm

03:08:35.520 --> 03:08:38.157
you know net making the world better
that I'm contributing to the people

03:08:38.240 --> 03:08:42.397
around me right and that I can look
back and say wow that was hard but it was

03:08:42.480 --> 03:08:46.397
worth it I think generally seems
to leave people in a better state than

03:08:46.480 --> 03:08:49.357
pursuit of pleasure pursuit
of quote unquote happiness

03:08:49.440 --> 03:08:52.797
>> does money have to do with They think
the founders and the founding fathers in

03:08:52.880 --> 03:08:55.917
the US threw this offkilter when they
used the phrase pursuit of happiness. I

03:08:56.000 --> 03:08:58.077
think they should have said
>> pursuit of satisfaction.

03:08:58.160 --> 03:09:00.717
>> They said pursuit of satisfaction.
We might live in a better world today.

03:09:00.800 --> 03:09:03.997
>> Well, you know, they could have
elaborated on a lot of things about

03:09:04.080 --> 03:09:05.437
[laughter] >> they could have
tweaked the second amendment.

03:09:05.520 --> 03:09:08.157
>> I think they were smarter than we
realized. They said, you know what,

03:09:08.240 --> 03:09:11.597
we're going to make it ambiguous and let
these uh these humans figure out the

03:09:11.680 --> 03:09:14.317
rest. These tribal cultlike humans

03:09:14.400 --> 03:09:16.797
figure out the rest.

03:09:16.880 --> 03:09:22.077
Uh but money empowers that. So I I think
and I think there I mean look I think

03:09:22.160 --> 03:09:24.637
Elon is I don't think I'm even a great
example but I think Elon would be the

03:09:24.720 --> 03:09:27.437
great example of this which is like
you know look he's a guy who from every

03:09:27.520 --> 03:09:29.917
every day of his life from the day
he started making money at all he just

03:09:30.000 --> 03:09:34.237
plows it into into the next thing.
Um and so I think I think money is

03:09:34.320 --> 03:09:37.357
definitely an enabler for satisfaction
this way money applied to happiness

03:09:37.440 --> 03:09:38.957
leads people down very dark paths.

03:09:39.040 --> 03:09:43.197
>> Yeah [laughter] very
destructive avenues. Uh money

03:09:43.280 --> 03:09:47.357
applied to satisfaction I think could
be is a real tool. Um, I always like, by

03:09:47.440 --> 03:09:50.477
the way, I always liked, you know, Elon
is the case study for behavior. But the

03:09:50.560 --> 03:09:53.837
other thing that it's always really made
me think is Larry, Larry Page was asked

03:09:53.920 --> 03:09:56.157
one time what his approach
to philanthropy was, and he said, "Oh, I'm

03:09:56.240 --> 03:10:01.197
just my my philanthropic plan is just
give all the money to Elon." [laughter]

03:10:01.280 --> 03:10:03.597
>> Right. >> Uh, well,
let me actually ask you about

03:10:03.680 --> 03:10:07.677
Elon. What What are your um You've
interacted with quite a lot of

03:10:07.760 --> 03:10:11.037
successful engineers and business people.
What do you think is special

03:10:11.120 --> 03:10:13.597
about Elon? We talked about Steve Jobs.

03:10:13.680 --> 03:10:16.317
what um what do you think is special

03:10:16.400 --> 03:10:18.397
about him as a leader as an innovator?

03:10:18.480 --> 03:10:22.717
>> Yeah, so the the core of it is he's
a he's he's back to the future. So he he

03:10:22.800 --> 03:10:26.077
is he is doing the most leading edge
things in the world but with an with a

03:10:26.160 --> 03:10:30.237
really deeply old school approach.
Um and so to find comparisons to Elon, you

03:10:30.320 --> 03:10:32.157
need to go to like Henry Ford and Thomas

03:10:32.240 --> 03:10:34.797
Watson and Howard Hughes and Andrew

03:10:34.880 --> 03:10:39.677
Carnegie, right? Um Leland Stanford,
[snorts] um John Dy Rockefeller, right?

03:10:39.760 --> 03:10:42.877
you need to go to the what were
called the bajgeoa capitalists like the

03:10:42.960 --> 03:10:46.797
hardcore business owner operators
who basically built you know indust

03:10:46.880 --> 03:10:49.437
basically built industrialized society

03:10:49.520 --> 03:10:53.037
um Vanderbilt um and it's a level of

03:10:53.120 --> 03:10:58.397
hands-on commitment um and uh depth um

03:10:58.480 --> 03:11:01.677
in the business um coupled with an

03:11:01.760 --> 03:11:05.197
absolute priority uh towards truth um

03:11:05.280 --> 03:11:08.397
and towards um how to put it science and

03:11:08.480 --> 03:11:13.037
technology uh town to first principles
that is just like absolute just just

03:11:13.120 --> 03:11:14.717
like unbelievably absolute. >> Mhm.

03:11:14.800 --> 03:11:18.077
>> He really is ideal day he's only ever
talking to engineers like he does not

03:11:18.160 --> 03:11:22.157
tolerate He has less tolerance
than anybody I've ever met.

03:11:22.240 --> 03:11:24.557
>> Um he wants ground
truth on every single

03:11:24.640 --> 03:11:28.797
topic. Um and he runs his businesses
directly dayto-day devoted to getting to

03:11:28.880 --> 03:11:30.797
ground truth in every single topic.

03:11:30.880 --> 03:11:34.797
>> So uh you think it was a good decision

03:11:34.880 --> 03:11:37.837
for him to buy Twitter? I have developed
a view in life to not second guessess

03:11:37.920 --> 03:11:42.557
Elon Musk. [laughter] >> I
know this is going to sound

03:11:42.640 --> 03:11:46.957
crazy and unfounded, but >> well
I mean uh he's got a quite a track

03:11:47.040 --> 03:11:50.637
record. >> I mean look, the car
was a crazy I mean the car was I mean look

03:11:50.720 --> 03:11:52.477
>> he's done a lot
of things that seem crazy.

03:11:52.560 --> 03:11:54.877
>> Starting a new car company in the
United States of America. The last time

03:11:54.960 --> 03:11:57.997
somebody really tried to do that was
the 1950s and it was called Tucker

03:11:58.080 --> 03:12:01.037
Automotive and it was such a disaster
they made a movie about what a disaster

03:12:01.120 --> 03:12:05.757
it was. Um and then rockets like who
does that? like that's there's obviously

03:12:05.840 --> 03:12:08.707
no way to start a new rocket company
like those days are over and then to

03:12:08.790 --> 03:12:12.237
[clears throat] do those at the same time.
So after he pulled those two off

03:12:12.320 --> 03:12:17.117
like okay fine like [laughter] like
this is one of my areas of like I whatever

03:12:17.200 --> 03:12:19.997
opinions I had about it that is just
like okay clearly are not relevant like

03:12:20.080 --> 03:12:22.797
this is you just you at some point
you just like bet on the person

03:12:22.880 --> 03:12:26.877
>> and in general I wish more people
would lean on celebrating and supporting

03:12:26.960 --> 03:12:28.957
versus deriding and destroying. Oh yeah.

03:12:29.040 --> 03:12:31.677
I mean, look, he drives resentment. Like

03:12:31.760 --> 03:12:33.917
it's resent. Like he he is a magnet for

03:12:34.000 --> 03:12:39.197
resentment. Um like his critics are the
most miserable like resentful people in

03:12:39.280 --> 03:12:43.517
the world. Like it's almost a perfect
match of like the most idealized, you

03:12:43.600 --> 03:12:47.117
know, technologist, you know,
of the century coupled with like just his

03:12:47.200 --> 03:12:50.557
critics are just bitter as can be.
I mean, it's it's I mean it's it's sort of

03:12:50.640 --> 03:12:53.277
very darkly comic to watch.

03:12:53.360 --> 03:12:56.157
>> Well, he uh he fuels
the fire of that by

03:12:56.240 --> 03:13:00.557
being an on Twitter at times.
And which is fascinating to watch the

03:13:00.640 --> 03:13:02.717
drama of human civilization given our

03:13:02.800 --> 03:13:07.117
cult roots just fully on fire.

03:13:07.200 --> 03:13:11.277
>> He's running a cult. [laughter] >>
You could say that >> very successfully.

03:13:11.360 --> 03:13:15.357
>> So now now that our cults have gone and
we search for meaning, what do you think

03:13:15.440 --> 03:13:16.717
is the meaning of this whole thing?

03:13:16.800 --> 03:13:18.717
What's the meaning of life, Mark Andre?

03:13:18.800 --> 03:13:23.917
>> I don't know the answer to that.
Um I think the meaning of uh of uh the

03:13:24.000 --> 03:13:27.037
closest I get to it is what I said about
satisfaction. So it's basically like

03:13:27.120 --> 03:13:30.317
okay we were given what we have
like we should basically do our best

03:13:30.400 --> 03:13:34.237
>> what's the role of love in that mix
>> I mean like what's the point of life if

03:13:34.320 --> 03:13:38.237
you're yeah without love like yeah
>> so love is a big part of that

03:13:38.320 --> 03:13:41.677
satisfaction and look like taking care
of people is like a wonderful thing like

03:13:41.760 --> 03:13:45.677
it you know a mentality you know there
are pathological forms of taking care of

03:13:45.760 --> 03:13:48.157
people but there's also a very
fundamental you know kind of aspect of

03:13:48.240 --> 03:13:50.957
taking care of people like for example
I happen to be somebody who believes that

03:13:51.040 --> 03:13:54.797
capitalism and taking care of people are
actually they're actually the same thing

03:13:54.880 --> 03:13:58.877
um somebody once that capitalism is how
you take care of people you don't know,

03:13:58.960 --> 03:14:02.717
>> right? Um, right. And so, like, yeah,
I think it's like deeply woven into the

03:14:02.800 --> 03:14:05.677
whole thing. Um, you know, there's
a long conversation to be had about that,

03:14:05.760 --> 03:14:08.877
but yeah. >> Yeah.
Creating products that are used by

03:14:08.960 --> 03:14:10.717
millions of people and bring them joy in

03:14:10.800 --> 03:14:13.037
smaller big ways. And then capitalism

03:14:13.120 --> 03:14:16.237
kind of enables that, encourages that.

03:14:16.320 --> 03:14:17.917
>> David Freriedman
says there's only three

03:14:18.000 --> 03:14:19.517
ways to get somebody to do something for

03:14:19.600 --> 03:14:23.637
somebody else. Love, money, and force.

03:14:23.720 --> 03:14:29.357
[laughter] And um love
and money are better than force.

03:14:29.440 --> 03:14:32.157
>> That's a good ordering. I think >>
we should we we should bet on those.

03:14:32.240 --> 03:14:34.797
>> Try love first.
If that doesn't work, then money

03:14:34.880 --> 03:14:38.957
>> and then force. Well, don't even try
that one. Uh Mark, you're an incredible

03:14:39.040 --> 03:14:42.397
person. I've been a huge fan. I'm glad
to finally got a chance to talk. I'm a

03:14:42.480 --> 03:14:46.557
fan of everything you do. Everything
you do, including on Twitter. It's a huge

03:14:46.640 --> 03:14:49.277
honor to meet you, to talk with you.
Uh thanks again for doing this.

03:14:49.360 --> 03:14:52.157
>> Awesome. Thank you, Lex.
>> Thanks for listening to this

03:14:52.240 --> 03:14:55.677
conversation with Mark Andre. To support
this podcast, please check out our

03:14:55.760 --> 03:14:58.557
sponsors in the description. And now,

03:14:58.640 --> 03:14:59.917
let me leave you with some words from

03:15:00.000 --> 03:15:03.357
Mark Andre himself. The world is a very

03:15:03.440 --> 03:15:05.677
malleable place. If you know what you

03:15:05.760 --> 03:15:08.317
want and you go for it with maximum

03:15:08.400 --> 03:15:11.517
energy and drive and passion, the world

03:15:11.600 --> 03:15:13.917
will often reconfigure itself around you

03:15:14.000 --> 03:15:16.237
much more quickly and easily than you

03:15:16.320 --> 03:15:23.240
would think. Thank you for listening
and hope to see you next time.